Healthcare Provider Absenteeism and Health Service Delivery in Kenya

category; binary indicator for primary care facilities Positive (+) Facility Location Urban location Binary indicator: 1 if facility in urban area; 0 if rural Positive (+) Facility Operations Hours facility is open Average daily operating hours (0-24 hours) Positive (+) Days open per week Number of days per week facility operates (1-7 days) Positive (+) Distance to sub- county HQ Kilometres from facility to sub-county headquarters; measures accessibility Negative (-) Facility Resources Medical staff Total medical staff employed at facility (doctors, clinical officers, nurses) Positive (+) Non-medical staff Total administrative and support staff employed at facility Positive (+) Equipment availability Binary indicator: 1 if facility possesses essential diagnostic and treatment equipment; 0 otherwise Positive (+) Electricity interruption Binary indicator: 1 if facility experiences frequent electricity interruptions; 0 otherwise Negative (-) Service Quality Diagnostic accuracy Proportion of diagnoses confirmed as accurate; measured on scale 0-1 Positive (+) Internet availability Binary indicator: 1 if facility has internet access; 0 otherwise Positive (+) Facility Size Medium or large facility Binary indicator: 1 if facility is medium or large; 0 if small Positive (+) Source: Author's computation 3.5 Data This study employs the 2018 Kenya Health Service Delivery Indicators survey data, collected from 3,094 health facilities nationwide. The dataset provides comprehensive representation across all 47 counties and includes both public and private facilities as well as all facility types: hospitals, health centres, and dispensaries. The KHSDI survey employed stratified random sampling with facility types and county location as stratification variables. Data were collected through unannounced facility visits, direct observation, and facility manager interviews. Healthcare worker absenteeism was assessed through observation of at least ten randomly selected healthcare personnel during unannounced visits, with absence rates calculated as the proportion of observed workers absent during normal facility operating hours. Health output data were obtained from facility records, ensuring objective measurement free from self-reporting bias. IV. Results and Discussion 4.1 Descriptive Statistics Table 2 presents descriptive statistics for variables employed in the regression analysis. The dataset comprises 2,684 health facilities with data on outpatient visits, inpatient bed-days, and facility deliveries. Healthcare provider absenteeism averages 0.714, indicating that 71.4 percent of observed workers were absent on average during facility visits, revealing a substantial provider absence challenge across Kenya's health system. Approximately 57 percent of facilities are publicly owned, while 27 percent are located in urban areas. Facility operational characteristics show that 27 percent operate over 18 hours daily, with 79 percent being dispensaries and clinics. Average diagnostic accuracy is 63 percent, and 26 percent of facilities have internet access. Mean medical staff per facility is 7.5, while non-medical staff average 5.0 employees. Approximately 89 percent of facilities are classified as small, reflecting Kenya's health system structure emphasising primary care accessibility. Table 2: Descriptive Statistics Variable Mean Standard Deviation Absenteeism 0.714 0.246 Ownership (1=public; 0=otherwise) 0.573 0.495 Location (1=urban; 0=otherwise) 0.265 0.441 Hours open per day Less than 8 hours 0.543 0.498 Between 9 and 18 hours 0.188 0.391 Over 18 hours 0.269 0.444 Facility type Hospital 0.052 0.223 Health centre 0.155 0.362 Dispensary and clinic 0.792 0.406 Diagnostic accuracy (%) 63.183 22.859 Facility has internet (1=yes; 0=otherwise) 0.259 0.438 Equipment availability (%) 72.67 14.896 Facility experienced electricity interruption 0.599 0.490 Days per week open 5 days 0.446 0.497 6 days 0.167 0.373 7 days 0.387 0.487 Time to sub-county headquarters 0.679 1.154 Number of medical staff in facility 7.531 18.585 Number of non-medical staff in facility 4.990 9.811 Facility size Small 0.889 0.314 Medium 0.087 0.282 Large 0.023 0.151 Source: Computed from Kenya Health Service Delivery Indicators 2018 survey data 4.2 Diagnostic Tests Results Prior to estimating the regression models, a series of standard diagnostic tests was conducted. Multicollinearity was assessed using variance inflation factors, which yielded a mean VIF of 2.812, well below the threshold of concern, with the highest individual value (8.569, for the dispensary/clinic category) remaining within acceptable range for categorical variables. Initial linearity checks based on residual versus fitted value plots indicated some departure from linear relationships, addressed by transforming dependent variables using a natural logarithm specification; resulting coefficients can be interpreted as elasticities. Residual diagnostics indicated that the normality assumption was reasonably satisfied based on histogram and quantile- quantile plot inspection; given the large sample size (N = 2,684), minor departures from normality are unlikely to materially affect inference. White's test for homoscedasticity indicated evidence of heteroscedasticity; accordingly, robust (heteroscedasticity-consistent) standard errors were used throughout. On the whole, these checks suggest that the estimated models satisfy key assumptions required for reliable regression-based inference. 4.3 Regression Results Table 3 presents results from the multiple linear regression models examining relationships between healthcare provider absenteeism and health service delivery outputs. Table 3: Regression Results Variable Outpatient Visits Inpatient Bed- Days Deliveries Absenteeism (proxied by absence rate) -311.853*** (138.00) -1934.704 (1425.463) 9.520*** (2.910) Ownership (1=Public; 0=Otherwise) 1158.356*** (88.200) -849.584 (650.584) 7.767*** (1.847) Location (1=Urban; 0=Otherwise) 143.008 (89.652) 764.481 (590.482) -3.273* (1.877) Hours Open 61.365 (59.760) -5.467 (617.301) 2.810** (1.251) Facility Type (Ref = First level hospital) Health Centre 998.153*** (204.357) 1,534.843** (757.357) 1.039 (4.283) Dispensary and Clinic 683.557*** (222.657) 1,971.136** (919.417) -8.603* (4.283) Diagnostic accuracy (%) 0.861 (1.411) 3.998 (11.518) -0.009 (0.030) Facility has internet (1=Yes; 0=No) 216.518*** (81.826) -836.092 (541.319) -3.198* (1.713) Equipment availability 10.575*** (66.702) 6.917 (13.156) 0.003 (0.033) Electricity interruption 210.176*** (66.702) 1098.005* (585.227) -1.018 (1.397) Days open 25.461 (59.083) -26.551 (794.033) 1.614 (1.237) Time to sub-county HQ -76.175*** (28.425) 96.343 (286.437) -0.058 (0.595) Medical staff 75.284*** (4.031) 91.613*** (14.218) 3.214*** (0.084) Non-medical staff -0.161 (5.248) 6.260 (18.912) -1.442*** (0.110) Facility Size (Ref = Small) Medium 343.908** (161.492) -540.660 (717.308) 12.873*** (3.382) Large 2,751.580*** (556.129) -2,539.627 (2,010.502) 77.220*** (11.646) Constant -1,461.874 (459.338) -1,926.399 (5330.878) -16.084* (9.619) R-squared 0.576 0.184 0.765 N 2,684 471 2,684 *p < 0.10; **p < 0.05; ***p < 0.01. Robust standard errors in parentheses. 4.4 Discussion of Results The results provide strong evidence that healthcare provider absenteeism constrains outpatient service delivery in Kenya. Facilities with higher absence rates serve markedly fewer outpatients, a pattern consistent with Tumlinson et al. (2019), who found through qualitative interviews with Kenyan providers that absenteeism increases patient waiting times and discourages future health seeking, and with Zhang (2021), who found in Uganda that rising absenteeism reduced the likelihood that patients sought care at public facilities and increased reliance on out-of-pocket treatment. The high average absence rate of 71.4 percent observed in this sample is broadly consistent with Chaudhury et al. (2006), who documented health worker absence rates of around 35 percent across several developing countries, and with a 25 percent absence rate recorded in an earlier Machakos District study in Kenya, suggesting that absenteeism remains a persistent and possibly worsening feature of the Kenyan health system. The relationship between absenteeism and inpatient bed-days was negative but not statistically significant. This is plausible given the smaller sample available for inpatient analysis (N = 471) and the relatively weak explanatory power of that model (R-squared = 0.184), consistent with the broader observation in the literature that absenteeism's effects appear to differ across service types and facility levels (Obodoechi et al., 2021). This may also reflect that facilities reorganise staff to maintain inpatient cover during absences, prioritising admitted patients over outpatient flow. The positive association between absenteeism and facility deliveries runs counter to expectations and to the general thrust of the literature, which links absenteeism to poorer maternal health outcomes. Goldstein et al. (2013), for example, found that nurse presence at a woman's first antenatal visit substantially increased the likelihood of facility-based delivery and improved related outcomes. Several explanations are plausible: facilities may continue to prioritise delivery services even when absenteeism affects other departments; absenteeism may be concentrated among non-obstetric staff; or the absence measure may not align well in timing with delivery records. Given that this result conflicts with the wider literature, it should be treated cautiously and flagged as an area requiring further investigation. Facility ownership, type, staffing, infrastructure, and size all show associations with service volumes broadly consistent with expectations and with the wider efficiency literature reviewed earlier. Public facilities and larger facilities serve substantially higher patient volumes; medical staffing is positively associated with service delivery across all three outcomes; and facilities further from sub-county headquarters serve fewer outpatients, echoing the access constraints emphasised by Tumlinson et al. (2013) and Mbombi et al. (2018). Service volumes are shaped jointly by workforce presence, facility scale, infrastructure, and geographic accessibility, with absenteeism representing one important but not isolated constraint among these factors. V. Conclusions, Limitations, and Policy Implications 5.1 Conclusion This analysis of 2,684 health facilities across Kenya's 47 counties confirms that healthcare provider absenteeism is widespread and is associated with meaningfully lower outpatient service utilisation, consistent with qualitative and quantitative evidence reviewed from Kenya and other developing country settings (Chaudhury et al., 2006; Tumlinson et al., 2019; Zhang, 2021). Beyond the direct reduction in service volumes, absenteeism likely undermines health system effectiveness more broadly by increasing waiting times, eroding patient trust, and shaping health-seeking behaviour in ways difficult to capture fully in facility-level data. The analysis also points to facility and organisational characteristics that shape service delivery capacity alongside absenteeism. Public ownership, adequate clinical staffing, larger facility size, equipment availability, and extended operating hours are all associated with greater service delivery, while geographic isolation from sub-county headquarters constrains access, particularly for outpatient services. The unexpected positive association between absenteeism and deliveries does not overturn this overall picture but suggests that the relationship between workforce presence and service delivery is not uniform across service types. Overall, the findings reinforce the view that workforce presence, alongside infrastructure and geographic accessibility, constitutes a fundamental constraint on health service delivery in Kenya. Addressing absenteeism represents a potentially high-return area for health system strengthening under the devolved county system. 5.2 Limitations The analysis is based on a single cross-section of facility-level data from the 2018 KHSDI (the newest KHSDI dataset currently available), which limits the ability to draw causal conclusions; the associations identified should be interpreted as correlational rather than as evidence of a direct causal effect, in line with the cautious interpretation adopted in the related literature (Obodoechi et al., 2021). The absenteeism measure is based on facility-level reporting and may be subject to measurement error or misreporting, which could partly explain the counterintuitive finding for deliveries. The inpatient bed-days model explains only a small share of variation, indicating that important determinants of inpatient utilisation are not captured by the variables available in this dataset. Finally, as a facility-level analysis, the study cannot speak directly to patient-level health- seeking behaviour or treatment decisions, which are the ultimate channels through which absenteeism is thought to matter (Tumlinson et al., 2019; Goldstein et al., 2013). 5.3 Policy Implications and Future Research Addressing healthcare worker absenteeism is essential for improving Kenya's health outcomes. Policies targeting workforce management, including improved supervision, accountability mechanisms, remuneration structures, and working conditions, are critical. Future work could draw on panel or longitudinal facility data, similar to the approach used by Zhang (2021) in Uganda, to examine how changes in absenteeism within facilities over time relate to changes in service delivery. Linking facility-level absenteeism data to patient or household-level surveys would help clarify whether the outpatient effects identified here translate into reduced healthcare seeking and worse health outcomes. The unexpected delivery result would benefit from qualitative follow-up, in the spirit of Tumlinson et al. (2019), to understand how staffing and absenteeism patterns specifically affect maternity units. Future studies could also examine the role of supervision and accountability mechanisms directly, building on Obodoechi et al. (2021), to assess which interventions are most effective at reducing absenteeism and improving service delivery within the Kenyan context. References Becker, G. S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. Columbia University Press. Bloom, D. E., & Canning, D. (2003). The health and wealth of nations. The Harvard Review of Medicine, 10, 1–15. Chaudhury, N., Hammer, J., Kremer, M., Muralidharan, K., & Rogers, F. H. (2006). Missing in action: Teacher and health worker absence in developing countries. Journal of Economic Perspectives, 20(1), 91–116. https://doi.org/10.1257/089533006776526058 Goldstein, M., Zivin, J. G., Habyarimana, J., Pop-Eleches, C., & Thirumurthy, H. (2013). The effect of absenteeism and clinic protocol on health outcomes: Evidence from Kenya. American Economic Journal: Applied Economics, 5(4), 46–78. https://doi.org/10.1257/app.5.4.46 Grossman, M. (1972). On the concept of health capital and the demand for health. Journal of Political Economy, 80(2), 223–255. Kenya Health Service Delivery Indicators. (2019). Kenya Health Service Delivery Indicators 2018 survey report. Kenya National Bureau of Statistics and Ministry of Health. Mbombi, M. O., Mavundla, T. R., & Botma, Y. (2018). The effect of absenteeism on the well- being and performance of nurses working at a tertiary hospital in Limpopo Province, South Africa. SAGE Open Nursing, 4, 1–10. https://doi.org/10.1177/2377960818779784 Mincer, J. (1974). Schooling, experience, and earnings. Columbia University Press. Mirrlees, J. A. (1974). Notes on welfare economics, information and uncertainty. In M. Balch, D. McFadden, & S. Wu (Eds.), Essays on economic behavior under uncertainty (pp. 243– 258). North Holland Publishing. Mushkin, S. J. (1962). Health as an investment. Journal of Political Economy, 70(5), 129–157. O'Brien, R. M. (2007). A caution regarding rules of thumb for variance inflation factors. Quality Engineering, 19(3), 198–213. https://doi.org/10.1080/08982110701451773 Obodoechi, D. N., Ndukwu, C. I., Ndukwu, E. E., & Nnadi, D. C. (2021). Health worker absenteeism and supervision in healthcare service delivery in Nigeria: A cross-sectional study. International Journal of Health Services Research and Policy, 6(2), 85–99. Ross, S. A. (1973). The economic theory of agency: The principal's problem. American Economic Review, 63(2), 134–139. Stiglitz, J. E. (1974). Incentives and risk sharing in sharecropping. Review of Economic Studies, 41(2), 219–255. Tumlinson, K., Gichane, M. W., Curtis, S. L., & LeMasters, K. (2019). Understanding healthcare provider absenteeism in Kenya: A qualitative analysis. Human Resources for Health, 17(1), 1–12. Tumlinson, K., Ye, Y., & LeMasters, K. (2013). Using mystery clients to assess provider adherence to national guidelines in Kenya. Studies in Family Planning, 44(3), 287–303. World Bank. (2015). World Bank country and lending groups. World Bank. World Health Organization. (2016). Global expenditure on health: Public spending on health at a glance. WHO. Wooldridge, J. M. (2012). Introductory econometrics: A modern approach (5th ed.). South- Western Cengage Learning. Zhang, H. (2021). Impact of health worker absenteeism on patient healthcare seeking behavior: Evidence from a longitudinal study in Uganda. The World Bank Economic Review, 35(2), 234–256.