Analysis of Performance, 2013/14

Transcription

1 REPUBLIC OF KENYA Ministry of Health Analysis of Performance, 2013/14 Transforming Health: Accelerating Attainment of Health Goals

2

3 Table of Contents List of tables and figures... iv Foreword... v Abbreviations... vi CHAPTER 1: BACKGROUND... 1 Introduction... 1 Data sources... 1 CHAPTER 2: OBSERVATIONS REGARDING SECTOR ACHIEVEMENTS... 3 Financing of health... 4 Investments in health... 5 Health outputs... 8 Health outcomes... 8 Health impact CHAPTER 3: ANALYSIS OF SECTOR PERFORMANCE Efficiency in production of health outcomes Equity in production of health outcomes Equity in access to services Equity in utilization of services CHAPTER 4: CONCLUSIONS AND RECOMMENDATIONS... 27

4 List of tables and figures Table 1: Contribution of health to GDP... 4 Table 2: Annual production accounts for health, Table 3: Trends in facility types, before and after the analysis period... 5 Table 4: Variables used for county efficiency analysis Table 5: Technical efficiency, by County Table 6: Scale efficiency, by County Figure 1: KHSSP Framework for Implementation... 3 Figure 2: Contribution of public and private sectors to health GDP... 4 Figure 3: Trends in facilities per 10,000 population per county at beginning, and end of the reporting period... 6 Figure 4: Trends in OPD per capita utilization by County, 2013/ Figure 5: Trends in Skilled Birth Attendances, Figure 6: Distribution of health outcomes, measured by an RMNCAH index Figure 7: Variations in deaths of 2013 versus Figure 8: Proportion of Counties at different levels of technical efficiency Figure 9: Beds per 1,000 vs poverty levels by county, Figure 10: Health workforce per 1,000 vs poverty levels by county, Figure 11: OPD per capita utilization vs poverty levels by county, Figure 12: RMNCAH index vs poverty levels by county, Figure 13: RMNCAH index vs poverty levels by county, Figure 14: Deliveries by skilled birth attendant s vs poverty levels by county, Figure 15: % children stunted vs poverty levels by county,

5 Foreword The health sector in Kenya is increasingly focusing investments to the population in a manner that ensures they receive services when needed, where needed and in a manner that addresses their needs. Such targeting of investments calls for innovative ways to better understand how these investments are made, and eventually used. As part of the health sector performance monitoring, performance reports that give one a fair picture of how the sector performed (based on our planned interventions) have been developed. Such performance assessments are good accountability tools that can be used to ensure we are reporting back to the citizens of the country on how we are utilizing their resources; and to make available for them the health services they require. This analytical report complements the annual performance review report (2013/2014). It is aimed at health sector actors, to provide a more critical, in-depth analysis the performance based on our focus and ideals that had been set out as a sector. As we provide services, we strive to ensure that these services are made available to the population as they need it, and is done in the most efficient and equitable manner. We attempt to provide through this report a more critical look at the kinds of activities we are prioritising and the impact they are having, as well as analyse how well we are being efficient and fair in distribution of our investments and achievement of outcomes. Based on this, we expect the sector will be able to draw strategic inferences regarding what they need to focus on to better respond to the needs of the people of Kenya. All Kenyans should be able to enjoy the right to health in a manner that doesn t discriminate against them and using scarce resources in the most efficient manner. This first analytical report of the health sector has been developed by the National Ministry of Health in line with its mandate to monitor and strategically guide health focus in the country. Its development has been carried out under my direct oversight, together with Dr. Isabella Maina head of the health sector Monitoring and Evaluation unit and with technical guidance from Dr. Humphrey Karamagi and Mr. Hillary Kipruto from the World Health Organization, plus Dr. Edwine Baraza, a post-doctoral researcher at the KEMRI Wellcome Trust Research Program. I urge all the Health Stakeholders to read the report take note of the lessons learnt and ensure the recommendations are implemented. In particular, the County Health management are encouraged to ensure that data management is strengthened and take cognizance of the disparities in equity and efficiency and endeavour to improve this which would eventually improve health outcomes and consequently the health of the Kenyan people. Dr. Nicholas Muraguri DIRECTOR OF MEDICAL SERVICES

6 Abbreviations AHSPR CDF CEC DHIS GDP FMS HIV KES LLITN MAL MFL NCD NTD RMNCAH SARAM SE TB TE US$ WHO Annual Health Sector Performance Report Constituency Development Fund County Executive Committee District Health Information System Gross Domestic Product Free Maternity Services Human Immunodeficiency Virus Kenya Shillings Long Lasting Insecticide Treated Nets Malaria Master Facility List Non Communicable Disease Neglected Tropical Disease Reproductive, Maternal, Newborn, Child, Adolescent Health Service Availability and Readiness Assessment Mapping Scale Efficiency Tuberculosis Technical Efficiency United States Dollars World Health Organization

7 CHAPTER 1: BACKGROUND Introduction The health sector in Kenya is one of the highly devolved sectors, implying service provision and management is now primarily a function of the counties. This calls for more in-depth and innovative ways of identifying areas of focus in order to attain the desired health goals not only for the counties (that are now responsible for service delivery) but also for the country at large. At present, the health sector has instituted a process to monitor sector performance. This focuses on monitoring the achievements against planned targets for different services and indicators, and is a critical management tool. However, there is still a lack of an analytical process that will generate the required knowledge to guide decision makers at national and county levels on the key actions they need to strategically focus on, if they are to achieve their health goals. The development of this analytical report is a step in this direction. It aims to provide a more critical analysis of the existing health data and information, to assist the sector draw strategic inferences regarding what they need to focus on to better respond to the needs of the people of Kenya. Data sources The report is based on analysis of the existing health sector information. It doesn t generate new information, but rather relies on whatever information exists relating to health, to draw the analysis and inferences needed to guide strategic health decision making. This of course means that it relies on the accuracy of the information as reported by the health sector. But, it is the view of the health sector that however inaccurate the information is, it still provides a sneak preview into what is going on in the health sector rather than having no information at all. Over time, as it is apparent what health information is used for, the quality of data generated for health will continue to improve. As the report is an analysis of health, not just of Ministry of Health, it draws on information from multiple sources, depending on the area of focus. The key information sources include: i) The Annual Health Sector Performance Report (AHSPF), July 2013 June 2014 from the Ministry of Health [1]that captures sector performance for the financial year under analysis ii) National mortality statistics from the Civil Registration Department that captures information on mortality trends iii) National Service Availability and Readiness Assessment Mapping 2014 from the Ministry of Health [2]that mapped the state of all health sector inputs at the beginning of the period of analysis iv) Kenya Economic Survey 2014 [3]by the Ministry of Planning that captures overall economic indicators and county poverty rankings for an equity analysis v) Assessment of the impact of Free Maternity Services (FMS) [4]by the Ministry of Health that documents progress, issues and challenges with the FMS policy implementation 1 P a g e

8 vi) Assessment of the status of service delivery in primary care facilities[5], and hospitals [6] by the Ministry of Health in 2014 vii) The Ministry of Health employee, work environment and client satisfaction survey of 2014 [7] This report is not able to provide all the different angles of analysis that can be carried out from the data in these various sources of information. It is, rather, an attempt to illustrate the depth of inferences that can be drawn from the existing health information, without the need for additional data. It is our hope that this effort will be continued in subsequent years, and this form of analysis deepened in scope and breadth to provide more detailed evidence for decision making in the health sector. The report is structured in two key formats. First it presents overall observations of health achievements at the financing, investment, output, outcome and impact levels, based on the existing data from the various sources. Following this, we introduce some analytical methods that can be used to understand, better, the sector information. For this particular report we focus on analysis of efficiency and equity in delivery of services using known tools. It is out hope that you find the reading of the report quite beneficial to you. 2 P a g e

9 HEALTH RESOURCES CHAPTER 2: OBSERVATIONS REGARDING SECTOR ACHIEVEMENTS Observations relating to health sector achievements are captured following the overall KHSSP framework, linking impact to financing (see below). Figure 1: KHSSP Framework for Implementation INPUTS/PROCESSES OUTPUTS OUTCOMES IMPACT Organization of Service Delivery Human Resources for Health Eliminate Communicable conditions Health Infrastructure Halt / reverse Non Communicable Diseases Health Products & Technologies ACCESS TO SERVICES Reduce violence and injuries BETTER HEALTH, IN RESPONSIVE MANNBER Health Information Health Leadership QUALITY OF CARE DEMAND FOR CARE Provide essential health care Health Financing Minimize risk factor exposure Health Research Strengthen cross sectoral collaboration This is based on the premise that: 1. Health resources are used to finance different inputs and processes needed to make the health system functional; 2. The levels and forms on investments in the health system are meant to attain desired improvements in health outputs, captured across improvements in access to services (physical, financial and socio-cultural), quality of care (client experiences, patient safety, effectiveness of care), and demand for services (healthy behaviours, health seeking behaviour); 3. Achievement of these outputs are what drive the improvements in coverage with KEPH services, measured by better utilization of health and related services across the six sector objectives; 4. These improvements in coverage with KEPH give the desired sector impact of better responsive health We reflect on the kinds of achievements made across each of these domains during the year under review. The information is from various sources, as highlighted in the previous chapter. 3 P a g e

10 Financing of health According to the 2014 Kenya Economic Survey report[8], the overall economy continued to grow during the FY 2013/14. Public revenues increased by 21.8%, from KES 831.1bn to KES 1,017.7bn. Public expenditure was at KES 1.3trillion, of which KES 395.4bn was on the social sector (a 7.1% increase from the previous year). The contribution of the health industry to the overall GDP is shown below. Table 1: Contribution of health to GDP OVERALL GDP at market prices 2,375,971 2,570,334 3,047,392 3,403,534 3,797,988 Total Health industry 60,196 64,738 74,237 81,850 72,914 Private actors contribution 33,525 34,920 38,805 42,153 45,112 Government contribution 26,671 29,818 35,432 39,697 27,803 Source: Economic Survey 2014, table 2.3 The contribution of the health industry to the overall GDP has remained low, and showed a reduction to 1.9% from 2.4%. The increases in GDP are benefitting other sectors relative to health. The proportional contribution of the private sector (corporations, households, and non-profit institutions) to the overall health GDP increased during the period of review. Figure 2: Contribution of public and private sectors to health GDP Source: Calculated from Economic Survey 2014 The trends in the annual production accounts for health in the country also show a reduction in the health output (expenditure), with the per capita health spending suggestive of a reduction from KES 3,046 (US$ 35.84) to KES 2,722 (US$ 32). 4 P a g e

11 Table 2: Annual production accounts for health, Variable Output at basic prices 86,468 91, , , ,791 Intermediate consumption 26,272 26,974 33,195 42,139 40,877 Gross value added at basic prices 60,196 64,738 74,237 81,850 72,914 Compensation of employees 40,904 44,107 51,983 58,355 47,590 Gross operating surplus / mixed income 19,293 20,630 22,254 23,495 25,324 Total health expenditure per capita (KES) 2, , , , , Total health expenditure per capita (US$) Employee compensation as % of total expenditure 47.3% 48.1% 48.4% 47.1% 41.8% Source: Economic Survey 2014, table 2.6 We see a marked reduction in the compensation of employees during 2013 (over 10billion KES), which appears to be the key driver in the reduction of the health expenditures. This could be a statistical artefact due to delays in incurring PE expenditures that occurred during the period. The increases occasioned by the other health factors of production are minimal, suggesting there is minimal increases (if any) in health financing during the period of our analysis. Investments in health The Service Availability and Readiness Assessment Mapping (SARAM) exercise provided the health sector with a comprehensive baseline regarding availability of services, and investments at the beginning of the reporting period. It is therefore possible to review the changes in investments from the time, to build a picture of what the sector constituents have been investing in. We start by reviewing the state of health infrastructure. We use the overall availability of functional facilities as captured in the SARAM exercise at the beginning of the reporting period, and from the Master Facility List at the end of the reporting period. The data suggests the health sector has been able to increase the number of functional health facilities (described as a facility that is providing some form of health services) by about 20%, with a current estimate of 9,642 health facilities being functional. The largest increase in facility types was from the private health facilities (36%), while faith based facilities showed the smallest increase (5%). Public facilities increased by only 15%. Table 3: Trends in facility types, before and after the analysis period Before After Variance % change Total number of facilities % Type of facility by Hospitals % level of care Primary care facilities % Facility by Public % ownership Faith based % Private % Source: SARAM 2013, and Annual sector performance report 2013/14 5 P a g e

12 The increases in facilities is not uniform across the country, with the largest increases seen in Nyeri, Kirinyaga, Kajiado and Kiambu counties. The trends, by county per 10,000 populations are shown below. Figure 3: Trends in facilities per 10,000 populations per county at beginning, and end of the reporting period By the end of the reporting period, the sector had 2.2 facilities per 10,000 persons, as compared to 1.8 facilities per 10,000 persons. This increase could be attributed to better reporting on facility functionality by counties, but is more likely a result of county efforts to make functional a number of health facilities they inherited that were either in the process of construction, or which were not yet functional. The CDF, local authorities, and other public sources of financing had been investing in health in the period prior to county functionality, which could explain the public facilities. On the other hand, the large increase in private facilities could point to better reporting of public health facilities in the Master Facility List. Looking at the health workforce, useable information is difficult to come by, particularly as a result of the transition of the health workforce management to counties during the period. However, we can infer that the disruptions in personnel emoluments for health workers noted during the year reduced their productivity. The effects on health services arising from this reduced productivity would be most marked in the counties that witnessed severe disruptions in services and less so in those counties that managed to maintain personnel emoluments for their staff. Additionally, this effect should be blunted by the as of now anecdotal evidence of increased availability of lower level cadres recruited by counties to make functional their lower level facilities. It is however imperative that both levels of government must put in order mechanisms of monitoring and reporting on the health workforce delivering health services both at the county and at the National level. Health products availability and use also is quite difficult to discern in the absence of county-wide information. Anecdotal evidence from Counties suggests there is increased investment in purchase of commodities due to local purchasing by counties to complement the nationally available budgets. 6 P a g e

13 However, information from the state of health services reports [5], [6] suggests a still weak picture on the ground. At the end of the reporting period, less than 50% of primary care facilities had all the tracer commodities in stock with 47.5% and 26.9% availability of the 16 tracer drugs in health centres and dispensaries respectively. Similarly, only 50% of health centres had all the 16 tracer nonpharmaceuticals in stock while only 19.2% of health centres were fully stocked indicating a major challenge for dispensaries since tracer commodities must be available in all facilities at all times for efficient delivery of quality services. Regarding expiries, there was a 15.0% and 11.5% of expired drugs in health centres and dispensaries respectively which was way higher than the recommended level of 5%. Regarding health financing, counties implemented, to different levels, the free maternity and user fees policies that the government enacted just prior to the reporting period. The assessment of the free maternity services showed the policy is implemented in various forms in all the counties and has reduced financial barriers to maternal services in the country. Efforts to scale up insurance mechanisms and other approaches to improve efficiency in use of resources however did not show any significant improvements during the year. The health information systems strengthening witnessed a number of challenges, with reduced reporting in the DHIS noted in the 1 st half of the year, which however improved towards the end of the reporting period due to direct advocacy efforts with counties to improve on reporting. In addition, the efforts to improve availability and use of vital statistics started to bear fruits, with the country production of the 1 st birth / death statistics of the past few years achieved. Coordination of health research through the newly established research unit in the MOH also showed improvements during the period under review. Efforts at improving service delivery systems were also noted during the year. The referral strategy launch, and various referral system strengthening initiatives were noted across counties. Most counties recognized the critical role of community systems and prioritized these, though only a few counties (Nakuru, Bungoma, Kilifi for example) made significant investments to improve availability of the comprehensive community care services. Anecdotal evidence suggests supportive supervision and mentoring systems reduced, with oversight and support to lower level facilities reducing in most counties. The KEPH was defined, but its operationalization not well disseminated to counties and facilities limiting its use in planning. Finally, there were no coordinated efforts discerned at monitoring / supporting the quality of facility based services. Finally looking at leadership and governance, we see the general trend in most of the counties being that of calling for quick results in health outcomes. To facilitate these, the health management teams in counties were largely left intact, with changes primarily at the political and administrative levels where CECs and Chief Officers were appointed. These provide the required political and administrative oversight of health activities as required in the constitution. We have, as a result, seen at least two trends in management structures emerging in the counties depending on how the technical health functions are managed 1. A single, versus multiple Directors (public health, clinical / medical) 7 P a g e

14 2. A single large county management team, versus a small county management team complemented by multiple sub county teams The technical utility of these different arrangements, in terms of their ability to deliver health results, needs to be further analysed to provide required guidance to counties on the most effective ways of managing the health technical functions. Health outputs As highlighted, it is expected that improvements in access, quality of care, and demand for services resulting from the various investments made in the sector. From this perspective, it is noted the sector primarily focused on improvements in access to services as opposed to quality of care, or demand for services. Such access improvements are most noted with physical access (more reported facilities / staff / commodities) and financial access (free maternity services, and free primary care services), though there is no clear evidence of improved socio-cultural access. The effects of this are quite varied across counties, as seen by the marked differences in OPD utilization across the counties (see figure below). Figure 4: Trends in OPD per capita utilization by County, 2013/14 Quality of care initiatives are mostly still at the drawing board, with very limited roll out across implementing units effected. The community based, and advocacy efforts to improve demand and use of available services were being scaled up in selected counties with no clear evidence these were having significant impact by the end of the reporting period. Health outcomes The result of this skewed focus on health outputs relating to access improvements is seen in the levels of health outcomes achieved. The most marked improvements in utilization are those that are most affected by improved access, while the least improvements are seen for those services reliant more in improvements in quality of care, or demand for services 8 P a g e

15 Figure 5: Trends in Skilled Birth Attendances, We also see a varied picture in terms of utilization of different health and related services across the counties, with the picture driven by the specific services a county is focusing on. There is no county witnessing a comprehensive and universal improvement in utilization across all the health outcome indicators, reflecting the skewing of investments towards specific services. We have derived an index from selected indicators purposively selected based on their focus on RMNCAH (the index being the average value for the indicators), to illustrate the differences across the counties. The indicators used to derive the index are the following: % Fully immunized infants % HIV + pregnant mothers receiving preventive ARV s % under 5 s treated for diarrhoea % School age children de-wormed reported in health facilities % Deliveries by skilled attendant in Health facilities to expected total deliveries (eligible pop.) % of Women of Reproductive age receiving Modern methods of Family planning % of newborns with normal birth weight % pregnant women attending at least 4 ANC visits 9 P a g e

16 The specific county achievement against this index is shown in the table below. Figure 6: Distribution of health outcomes, measured by an RMNCAH index We see a range from 20% in Mandera County, to 59% in Kericho County across the index highlighting the varied levels of utilization across the counties. These variations in outcomes are seen across many services, and so represent real differences in access, quality and demand for care across the country. Health impact Information on impact of the actions in the year are difficult to discern, as the data is usually collected through surveys. However with routine information, from civil registration (overall deaths registered), and the client satisfaction surveys. The Demographic and Health Survey that would provide information on age-specific mortality cannot be used, as - The DHS statistics are an average for a period of time, not only the year under consideration (for example, Maternal Mortality Rates are average for the preceding 10 years) - The DHS data was not completed, at the time of completion of this report. The civil registration data on deaths is not complete. However, it represents the only available and comprehensive data source, particularly when analysing trends over time where systematic errors in the data should be the same across years. Comparison of the deaths per 1,000 persons by Counties for 2012 and 2013 shows significant variations across Counties. Nationally, there were 4.44 deaths per 1,000 persons in 2014 as compared to 4.3 deaths per 1,000 persons in 2013, representing a minor increase in deaths by 0.15 per 1,000 persons. 22 counties registered reductions in overall registered deaths, with 25 counties registered increases. The 10 P a g e

17 county with the highest increase in deaths registered is Garissa County (2.19 more deaths per 1,000 persons), while Bungoma County registered the largest reduction in deaths registered (2.17 less deaths per 1,000 persons). There is therefore no any hard evidence of reductions in overall mortality. However, we see some rudimentary evidence of changes in geographical mortalities, though there is no specific discernible pattern (e.g. counties with more donor support having larger reductions in deaths). Figure 7: Variations in deaths of 2013 versus 2012 However it is expected that there shall be changes in the mortality patterns arising from the achievements reported by the sector. Given the improvements in access noted, it is expected that mortality due to conditions associated with urgent care (communicable diseases, acute events, or conditions in mothers, children or neonates) to have reduced. The reduction however is affected by the counter effect of non-improvement in quality of care, reducing its potential effect on overall mortality. Responsiveness of the services to the legitimate needs of clients represents one of the key impact thrusts of the sector. From the employee, client and work environment survey, overall patient satisfaction appears fair, with at least 63% of people reporting good access and 69% satisfied with the services (highest satisfaction being with immunization, while emergency and mortuary services had least satisfaction levels). 11 P a g e

18 CHAPTER 3: ANALYSIS OF SECTOR PERFORMANCE In this chapter, we analyse the sector performance in the following ways: 1. We look at the technical efficiency with which available inputs are used to produce the desired outcomes. By analysing the technical efficiency in production of health, we are able to provide guidance on how much more outputs can be produced with the given investments, or looking at it the other way we can tell by how much we can reduce the current investments, without affecting the achieved outcomes. We therefore are able to make best use of the resources available. 2. We look at the equity / distribution of investments to better understand the variations in capacities to produce outcomes. We recognize some areas of the country are disadvantaged and so an analysis that compares investments across different equity levels allows us see how fairly investments are being made. Efficiency in production of health outcomes The health sector is unique, in that it requires multiple types of inputs / investments to produce a multiple set of outputs. A simple technical efficiency analysis therefore is difficult to conduct, as it is arguable which input / output best represents the health sector effort. Attempts have, however, been made with various methodologies to capture the multiple input / output nature of health sector actions, to allow for efficiency analyses. Efficiency is calculated relative a frontier function using either non-parametric mathematical programming methods or econometric/regression methods. Each have advantages / disadvantages, but the non-parametric mathematical programming method the Data Envelopment Analysis (DEA) is the most commonly used in health[9]. Its main advantage is that it is able to deal with multiple inputs and multiple outputs or services. It not only identifies inefficiencies, but also permits analysis of sources of inefficiency and quantification of magnitudes of inefficiencies in the production of outputs. It is for these reasons that, we considered DEA appropriate for the purposes of this study. For an appropriate technical efficiency analysis, we require a set of inputs, and outputs that are managed by, and the result of actions by a given Decision Making Unit (DMU). Such a Decision Making Unit should have a level of autonomy over its decision making process. We use each County as a standalone DMU as it possesses the decision making autonomy required for it to be a DMU. Our efficiency analysis is therefore based on comparison of the 47 Counties. We analyse the production by 12 P a g e

19 counties of the maximum possible outputs from its available investments. Being a non-parametric method of efficiency analysis, we compare counties against each other, with efficiency of a county being a measure relative to other counties. A county is rated as efficient if the other counties cannot show evidence of use of inputs in a better way. Its inputs (or outputs) cannot be improved without worsening some other inputs (or outputs). The DEA methodology will give us a number of efficient Counties, plus levels of variation from these efficient counties of the others. Given the multiple input / output nature of the health sector, there are various mixes of inputs / outputs that are considered efficient. The set of efficient counties form the frontier of efficiency, against which the inefficient counties will vary. From the method, we are able to derive two forms of efficiency: technical efficiency, and scale efficiency. Technical efficiency looks at the level of efficiency of a county that cannot be attributed to deviations from the optimal scale (where there are constant returns to scale one unit of input gives one unit of output). On the other hand, scale efficiency looks at the extent to which a county deviates from the optimal scale (one unit of input produces less or more than one unit of output). The technical efficiency is therefore able to tell us by how much we can reduce investments for a given level of output (or increase the outputs for a given level of investment), while scale efficiency is able to tell us by how much any additional investment will give us the desired level of outputs. Given the multiple inputs/multiple output nature of health, it is important to agree on a few inputs / outputs that will be used for the efficiency analysis. We focused on those indicators which are able to give a wide range of impacts across the sector, and have fair data across the counties. The input data was for the beginning of the reporting period (June 2013 sourced from the AHSPR, SARAM and MFL), while the output data was for the end of the reporting period (July 2014 sourced from the AHSPR, and DHIS). The variables used, together with the standard deviations of the data across the counties are shown below. Table 4: Variables used for county efficiency analysis Standard Deviation INPUT VARIABLES Number of hospitals per county Number of primary healthcare facilities per county Monetary allocations per county (KES) KES 1,618,565, (USD ) Number of healthcare workers per county 1, OUTPUT VARIABLES % Deliveries conducted by skilled attendant in Health facilities % pregnant women attending at least 4 ANC visits 9.82 % Infants under 6 Months on exclusive breastfeeding 2013/ Child Immunization coverage % new outpatient with high blood pressure 1.15 % New outpatients with Diabetes 0.30 outpatient annual utilization (per capita) rate 0.61 Source: Input data from SARAM / MFL, output data from DHIS / AHSPR 13 P a g e

20 The input variables were chosen to represent a set of different investments, for which good information exists across all the counties. All the input data was standardized to a per capita value, to weight the different populations on counties into the measure. Number of facilities (hospitals / primary care facilities) is a good input variable as it illustrates the level of physical access to services. The data used is for all facilities in the county public and non-public as these are all recognized service provision points. The counties with more facilities per person have more weight for this variable. The monetary allocation looks at the budgeted amount of funds allocated by treasury, to each county. This represents the potential amount of funds directly available to a county team to provide services. It is useful, as it also indirectly captures many other input investments (operational funds, medicines and supplies, etc). Different counties allocate different amounts to health, with the counties allocating more of this allocation to health being more advantaged with this input, as compared to those that use less of this available budget. The values used are only for the on-budget allocations, and don t include off budget amounts available to counties, which are difficult to determine in a similar manner, and most of these funds are not under direct control of the county teams their utilization is determined by the implementing partners. The number of health workers looks at the total available health workforce in the county both public and non-public. The counties with more HWs have more weight for this variable. On the other hand, the output variables used were aimed at having values that capture the wide scope of outcomes that health is attempting to attain, focusing only on indicators which have fairly good data across all the counties. Two maternal health indicators were used, given the strong sector focus on improving services to mothers deliveries conducted by skilled birth attendants and ANC 4 visits. The percentage utilization across counties was used, with counties that had higher utilization scoring higher. The analysis used two indicators addressing child health one for early childhood survival (exclusive breastfeeding), and the other for fully immunized children. Again, percentage utilization across counties was used, with counties that had higher utilization scoring higher. The analysis included two indicators for Non Communicable Diseases (NCDs) new outpatients with high Blood Pressure (BP), and diabetes to ensure this increasing health threat is included in the outcome analysis. As higher values denote lower success in a county at managing these risks, the analysis scored counties with higher values lower. Finally, the analysis included one general output indicator per capita new OPD cases to include in the weighting most of the other health services that could not be included due to data gaps The technical efficiency scores for the various counties are therefore shown in the table below. 14 P a g e

21 Table 5: Technical efficiency, by County RELATIVE TECHNICALLY EFFICIENT COUNTIES RELATIVE TECHNICALLY INEFFICIENT COUNTIES County name Efficiency score County name Efficiency score Baringo 100 Turkana 93.3 Busia 100 Tharaka Nithi Elgeyo Marakwet 100 Taita Taveta 61 Embu 100 Bomet Homa Bay 100 Kirinyaga Isiolo 100 Kisii Kiambu 100 Garissa Kilifi 100 Laikipia 30.3 Kisumu 100 Siaya 29.9 Kwale 100 Mombasa Lamu 100 Mandera Marsabit 100 Trans Nzoia Migori 100 Nandi Nairobi 100 Vihiga Narok 100 Nyamira Nyeri 100 Kericho 13.2 Samburu 100 Kajiado Tana River 100 Muranga 9.84 Wajir 100 Uasin Gishu 8.76 West Pokot 100 Nyandarua 8.26 Makueni 7.22 Nakuru 5.68 Machakos 5.04 Kitui 4.33 Bungoma 3.16 Meru 3.04 Kakamega 2.72 The mean scores of pure TE and SE of the counties were 56.43% (SD 41.64) and 50.25% (SD 36.81), respectively. Of the 47 county health systems, 20 (43 %) were technically efficient constituting the best practice frontier. The remaining 57 % were technically inefficient, with an average TE score of % (SD 22.90). This finding implies that these 27 inefficient counties could potentially produce % more outputs by utilizing the current levels of inputs (They could also reduce their current input endowment by % while leaving their output levels unchanged). A significant proportion (40%) of counties had a technical efficiency score of less than 30%, highlighting the significant variations in capacities to efficiently utilize available resources in the country (see figure below). There is a significantly wide variation in the use of available resources to produce the desired health outcomes. These counties with high levels of inefficiencies can therefore be focused on, to produce more health outcomes (or, their investments reduced without affecting their current outcomes). The sector has significant scope for improving on its health outcomes, therefore, by focusing on these inefficient counties to improve on health targets. 15 P a g e

22 Figure 8: Proportion of Counties at different levels of technical efficiency Additional efficiency analysis looked at scale efficiency, with the results shown in the next table. Table 6: Scale efficiency, by County County name Efficiency RTS Sum Lambda County name Efficiency RTS Sum Lambda Isiolo 100 CRS 1 Marsabit DRS 1.1 Lamu 100 CRS 1 Kirinyaga 34.1 DRS 1.16 Vihiga 100 CRS 1 Siaya DRS 1.11 Kitui CRS 1 Embu DRS 2.44 Kajiado CRS 1 Bomet DRS 1 Makueni CRS 1 Garissa DRS 1.15 Meru CRS 1 Mombasa DRS 1.16 UasinGishu CRS 1 ElgeyoMarakwet DRS 1.01 Muranga CRS 1 Busia DRS 1.33 Machakos CRS 1 Kwale DRS 1.11 Laikipia CRS 1 West Pokot DRS 1.01 Nyandarua CRS 1 Turkana DRS 1.01 Bungoma CRS 1 Baringo DRS 1 Nyamira CRS 1 Wajir DRS 1.03 Kakamega CRS 1 Kisii DRS 1.22 TaitaTaveta DRS 1.01 Nyeri DRS 1.18 Kericho CRS 1 Narok DRS 1 Tana River DRS 1 Migori 8.6 DRS 1.38 Nandi DRS 1 Kisumu 7.99 DRS 1.27 Samburu DRS 1 Homa Bay 5.22 DRS 1.05 TharakaNithi DRS 1.03 Kilifi 4.98 DRS 1.3 Nakuru DRS 1.06 Kiambu 4.06 DRS 1.68 Trans Nzoia DRS 1.13 Nairobi 1.08 DRS 1.21 Mandera DRS 1 16 P a g e

23 Only 3 (6 %) county health systems had an SE of 100%, implying thereby that they had the most productive scale size (MPSS) for that particular input-output mix. The remaining 44 (94 %) county health systems were found to be scale inefficient, manifesting a mean SE score of % (SD 35.57). This implies that, on average, the scale-inefficient county health systems could increase their output size by % at the current level of inputs. Alternatively, on average, the scale-inefficient county health systems could reduce their input size by % without affecting their current output levels. Of the Scale inefficient counties, 31 (70%) had decreasing returns to scale (DRS) implying that these inefficient counties need to scale down their operations to achieve constant returns to scale (CRS). 17 P a g e

24 Equity in production of health outcomes The nature of health services means it is a right for people to have access to, and use them. A comparison of the access and utilization of available health and related services is important in understanding the fairness of the levels of health investments. Regular data on use of services by the poor across counties during the period of the analysis is not directly available. However, we use comparison data to probe into equity in access, and use of services. We look at the available data by county in 2 key areas: 1. A comparison based on poverty levels of different investments in health as a probe into access, 2. A comparison across counties based on poverty levels of utilization of services Poverty levels by county are derived from the Economic survey 2014, while the rest of the data is from the AHSPR 2013/14. This analysis does not account for intra-county variations in poverty levels, which may be significant on their own. In the future, it will be critical for such analyses to drill down within the county for this kind of information. Equity in access to services We look at information on two key investment variables; infrastructure and health workforce. Beds (and cots) per 1,000 in each county, compared against the poverty levels by county are shown in the figure below Figure 9: Beds per 1,000 vs poverty levels by county, P a g e

25 Beds/cots per 1, Isiolo Lamu Nairobi Tana River Taita Taveta Embu Kajiado Kiambu Nyeri Mombasa Kisumu Tharaka Nithi Migori Elgeyo Marakwet Machakos Kisii Busia Samburu Marsabit Kirinyaga Meru Narok Nakuru Garissa Uasin Gishu Homa Bay Kericho Nyamira Vihiga Nyandarua Kakamega Laikipia Bomet Baringo Kitui Makueni Kilifi Kwale Wajir Muranga Siaya Nandi Trans Nzoia Bungoma West Pokot Mandera Turkana Poverty headcount(%) Of the counties with high levels of poverty, Isiolo has the highest investments in beds per 1,000 population, while the counties of Turkana, Mandera, Marsabit and Wajir are investing least in infrastructure. Additional infrastructure investments in these counties are needed, to improve access to services. Lamu and Nairobi appear to invest more than they may require in terms of infrastructure. It is important also to note that this analysis does not take into account the geographical distribution of the beds within the county which could be a significant factor. A further look at available health workforce per 1,000 persons in each county, compared against the poverty levels by county are shown in the figure below Figure 10: Health workforce per 1,000 vs poverty levels by county, P a g e

26 HWs per 1000 persons Embu Mombasa Uasin Gishu Isiolo Laikipia Tharaka Nithi Nairobi Kiambu Kirinyaga Meru Nakuru Garissa Nyandarua Taita Taveta Baringo Machakos Busia Samburu West Pokot Marsabit Kajiado Nyeri Lamu Siaya Muranga Narok NandiTrans Nzoia Elgeyo Marakwet Kitui Kericho Vihiga Kakamega Homa Bay Makueni Migori Bomet Nyamira Kisumu Bungoma Kisii Kilifi Kwale Tana River Wajir Mandera Turkana Poverty headcount(%) Again, we see the Northern Arid Lands having fewer investments in health workforce as compared to the levels of poverty. On the other hand, Embu, Mombasa and Uasin Gishu are outliers, with relatively high numbers of health personnel as compared to their poverty levels. For better equity in distribution of health workers, more staff need to be made available in these counties of Mandera, Turkana, Wajir, and Marsabit. Equity in utilization of services This perspective of looking at equity now compares health outcomes, with poverty headcounts. We would expect to have higher utilization of services with a higher poverty headcount, as we expect more poor persons to require more health services. The patterns for select outcome indicators are now discussed. We start by comparing the health outputs resulting from the investments in health, with the poverty headcount. The information, by County, is shown in the figure below. We again see the same pattern with health investments reflected at this output level, with per capita OPD utilization too low when compared with expected levels in the Northern Arid Counties of Turkana, Mandera, Wajir and Marsabit. The levels in Tana River also appear low compared to the investment level, and need further analysis. Figure 11: OPD per capita utilization vs poverty levels by county, P a g e

27 OPD utilisation rate Embu Nyeri Trans Nzoia Kirinyaga Isiolo Narok Tharaka Nithi Kilifi Taita Elgeyo Baringo Taveta Marakwet Lamu Kericho Muranga Siaya Laikipia Kiambu Nakuru Nyandarua Nandi Bomet Nyamira Machakos Kitui Makueni Kajiado Meru Migori Uasin Kakamega Gishu Kisii Busia Vihiga Homa Bay Mombasa Kisumu Samburu Kwale Nairobi West Pokot Garissa Bungoma Marsabit Tana River Wajir Turkana Mandera Poverty headcount(%) On the other hand, Embu County is clearly an outlier, with a significantly higher OPD per capita utilization as compared to all the other counties. The county did have a high health workforce count, but the infrastructure was not significantly higher than its peer counties. It should be recalled that the County was one of the frontier counties for technical efficiency, but scored poorly on scale efficiency and had decreasing returns to scale. A more in-depth analysis is needed, to better understand the dynamics occurring in Embu County. Looking at actual health outcomes, we present information in a variety of ways. First, we look at the RMNCAH index achievement by county, versus the poverty headcount. We would expect counties with higher poverty levels to utilize more RMNCAH services. The county comparisons are shown in the proceeding figure. We see the counties with low levels of poverty performing well in RMNCAH a very good sign that these counties are taking advantage of their existing investments to actually utilize available services. Of the Northern Arid Land counties, Garissa appears the outlier, being able to achieve RMNCAH outcomes higher than its peers. Figure 12: RMNCAH index vs poverty levels by county, P a g e

28 RMNCAH Index (%) Kericho Migori Busia Kilifi Kajiado Lamu Kisumu Siaya Taita Taveta Machakos Kirinyaga Kiambu Embu Uasin Gishu Homa Bay Muranga Makueni Nairobi Nyandarua Nandi Laikipia Tharaka Nithi Kisii Kitui Meru Elgeyo Marakwet Nakuru Nyeri Vihiga Nyamira Baringo Mombasa Kakamega Isiolo Kwale Narok Bomet Trans Nzoia Garissa Bungoma Samburu Tana River West Pokot Marsabit Wajir Turkana Mandera Poverty headcount(%) A further analysis of health outcomes compared across counties based on poverty headcount looked at skilled birth attendance, HIV+ve mothers receiving ARVs and stunting levels. Regarding HIV +ve mothers receiving ARVs, the county comparisons are shown in the figure below Figure 13: % HIV+ve mothers on ARVs vs poverty levels by county, 2014 Kirinyaga Siaya Kisumu Nairobi Muranga Makueni Nyandarua Vihiga Homa Migori Bay Nandi Machakos Taita Taveta Embu Kakamega Kisii Nakuru Nyamira Kwale Kajiado Laikipia Baringo Lamu Kiambu Meru Nyeri Garissa Mombasa Bomet Tharaka Nithi Kericho Trans Nzoia Kitui Busia Samburu Tana River Bungoma West Pokot Kilifi Marsabit Narok Uasin Gishu Isiolo Turkana Elgeyo Marakwet Mandera Wajir Poverty headcount(%) 22 P a g e

29 We see the same NAL Counties performing poorly equitably, while we see Isiolo, Elgeyo Marakwet also performing poorly when looking at equity in utilization. We also see Kajiado, Muranga, Nairobi and Kirinyaga doing well all counties with high levels of affluence. However, Siaya interestingly does well in ensuring their coverage of HIV+ve mothers are receiving ARVs as compared to its levels of poverty. The comparisons for deliveries by skilled birth attendants is shown below Figure 14: Deliveries by skilled birth attendants vs. poverty levels by county, 2014 Kiambu Migori Nairobi Kirinyaga Kisumu Kisii Nyeri Mombasa Siaya Nakuru Homa Bay Lamu Laikipia Taita Taveta Nyamira Uasin Gishu Kericho Elgeyo Marakwet Busia Kilifi Isiolo Kwale Kajiado Meru Muranga Narok Embu Nyandarua Vihiga Kakamega Tharaka Nithi Bungoma Garissa Bomet Nandi Trans Nzoia Baringo Machakos Kitui Makueni Marsabit West Pokot Samburu Tana River Wajir Turkana Mandera Poverty headcount(%) The NAL counties of Mandera, Turkana, Wajir and Marsabit again show very low levels of deliveries, compared to their high poverty headcount. On the other hand, Kiambu county is interesting, with a high level of deliveries as compared to its poverty headcount. This suggests more issues are at play that are leading to higher utilization of services. The final variable we analyse is for levels of stunting. This represents a more long term indicator of wellbeing, with high levels suggestive of long term malnutrition prevalent in the county. Comparing levels of stunting with poverty levels should show increasing stunting with increasing levels of poverty. The county picture is shown in the proceeding figure. The pattern we see is slightly different, from the health outcome data we have been showing. Turkana clearly has lower levels of stunting than would be expected given its poverty levels. We also see Kitui and Laikipia counties having higher than expected levels of stunting, suggesting some issues are going on there that need further investigation. On the other hand, Kisii, Elgeyo Marakwet, Trans Nzoia and Nyandarua counties all have lower than expected levels of stunting, also suggesting some events are taking place in these counties that are giving them better results than what would be expected for their levels of poverty. 23 P a g e