Signal functions for measuring the ability of health facilities to provide abortion services: An illustrative analysis using a health facility census in Zambia

Background: Annually, around 44 million abortions are induced worldwide. Safe termination of pregnancy (TOP) services can reduce maternal mortality, but induced abortion is illegal or severely restricted in many countries. All abortions, particularly unsafe induced abortions, may require post-abortion care (PAC) services to treat complications and prevent future unwanted pregnancy. We used a signal-function approach to look at abortion care services and illustrated its utility with secondary data from Zambia. Methods: We refined signal functions for basic and comprehensive TOP and PAC services, including family planning (FP), and assessed functions currently being collected via multi-country facility surveys. We then used the 2005 Zambian Health Facility Census to estimate the proportion of 1369 health facilities that could provide TOP and PAC services under three scenarios. We linked facility and population data, and calculated the proportion of the Zambian population within reach of such services. Results: Relevant signal functions are already collected in five facility assessment tools. In Zambia, 30 % of facilities could potentially offer basic TOP services, 3.7 % comprehensive TOP services, 2.6 % basic PAC services, and 0.3 % comprehensive PAC services (four facilities). Capability was highest in hospitals, except for FP functions. Nearly two-thirds of Zambians lived within 15 km of a facility theoretically capable of providing basic TOP, and one-third within 15 km of comprehensive TOP services. However, requiring three doctors for non-emergency TOP, as per Zambian law, reduced potential access to TOP services to 30 % of the population. One-quarter lived within 15 km of basic PAC and 13 % of comprehensive PAC services. In a scenario not requiring FP functions, one-half and one-third of the population were within reach of basic and comprehensive PAC respectively. There were huge urban-rural disparities in access to abortion care services. Comprehensive PAC services were virtually unavailable to the rural population. Conclusions: Secondary data from facility assessments can highlight gaps in abortion service provision and coverage, but it is necessary to consider TOP and PAC separately. This approach, especially when combined with population data using geographic coordinates, can also be used to model the impact of various policy scenarios on access, such as requiring three medical doctors for non-emergency TOP. Data collection instruments could be improved with minor modifications and used for multi-country comparisons.

We defined two levels of TOP and PAC services: basic and comprehensive, corresponding typically to hospitals and health centre services. Table 1 presents our criteria compared to the signal functions proposed for SAC by Healy et al. [8]. Signal function classification system: criteria for termination of pregnancy (TOP) and post-abortion care (PAC) in comparison to previously suggested SAC criteria comp comprehensive, iv intravenous, IUD intrauterine device aCriteria for safe abortion care (SAC) as defined by Healy and colleagues [8] are shown for comparison. In their classification system, staffing is implied by having service provision 24/7 but not stated explicitly Table 2 shows whether pertinent data were collected in five health facility data collection efforts identified in our previous work [24] and by others [27, 28]. Three other potential data collection tools (Rapid Health Facility Assessment, Quick Investigation of Quality, and Service Availability Mapping) were also examined, but discarded because they did not contain enough relevant aspects of TOP, PAC or EmOC, or of FP [29–31]. Table 2 also shows how our criteria were operationalized using data from the 2005 Zambia Health Facility Census (HFC). The HFC, a national-level assessment of health system assets, was developed by the Japan International Cooperation Agency [32] and covered 1421 facilities in Zambia, comprising all public and semi-public facilities (e.g., mission or non-governmental organizations), and some larger private-for-profit facilities. Data collection was done face-to-face by trained members from Zambian District Health Management Teams, supervised by personnel from the Provincial Health Office and by a National Steering Committee. It comprised questionnaires on infrastructure, utilities, equipment, service delivery and human resources, taking one or more days per facility to complete [33, 34]. Availability of items in selected health-facility assessment instruments, reviewed April 2015, and operationalization in Zambia Bracketed responses, e.g. (yes), signify there are caveats to the response, as indicated in the footnotes Zambia measured capability to provide different services as described in the text, mainly by interviewing key informants in facilities, checking inventories, and reviewing records 1 only asked if facility does: deliveries (AMDD or HFC); deliveries or newborn care (SPA); maternal health services (FASC); normal delivery or BEmOC or CEmOC or newborn care services (SARA) 2 only asked if facility does normal delivery, asks if removal of retained products performed after delivery 3 only asked if facility does delivery, asks availability of vacuum aspirator equipment or D&C kit 4 only asked if facility is hospital that offers surgical services (including minor surgery such as suturing, circumcision, wound debridement, etc.) or caesarean section 5 only asked if can treat haemorrhage, not specifically give uterotonics; availability of uterotonic stocks assessed 6 only asked if uses misoprostol to remove retained products 7 caesarean (FASC, HFC); caesarean & minor procedures (SPA); obstetric surgery e.g. caesarean (also asks about operating theatre even if no deliveries) (AMDD) 8 asked if provide temporary FP methods (pills, condoms, injectables, implants, & IUDs) all merged in one response 9 asked if provides D&E to remove retained products aasked for availability of misoprostol tablets, and of emergency contraceptive methods (e.g. levonorgestrel, ulipristal acetate, mifepristone) (merged together); if has FP services, asks if it provided emergency contraceptive services (e.g. levonorgestrel, ulipristal acetate, mifepristone) bonly asked if uses vacuum aspiration to remove retained products conly asked about case management for severe pneumonia and severe dehydration for children dasked hours not days open unless facility does caesarean (when 24/7 opening is assessed) AMDD averting maternal death and disability needs assessment toolkit. Available at: https://www.mailman.columbia.edu/research/averting-maternal-death-and-disability-amdd/toolkit. Accessed: 2015-04-18 SPA MEASURE DHS Service Provision Assessment (SPA). Updated June 2012. Available at: http://www.measuredhs.com/What-We-Do/Survey-Types/SPA.cfm. Accessed: 2015-04-10 SARA World Health Organization Service availability and readiness assessment. Version 2.2 December 2014. Available at: http://www.who.int/healthinfo/systems/sara_reference_manual/en/Accessed: 2015-04-18 FASQ MEASURE Evaluation’s Facility Audit of Service Quality. Version 1. 25 Nov 2008. Available at MEASURE Evaluation IHFAN (International Health Facility Assessment Network): http://ihfan.org. DDI TEMP FASQ 2008 v01. Accessed: 2015-04-26 HFC Japan International Cooperation Agency Health Facility Census. Available at MEASURE Evaluation IHFAN (International Health Facility Assessment Network): http://ihfan.org. Zambia 2008 HFC TEMP 2008-v02. Accessed: 2015-04-26 We then used the Zambia HFC to evaluate abortion service availability and coverage using our proposed signal functions. Data to determine PAC and TOP services came from two sets of questions: one on provision of FP and PAC services, asked of all facilities, and the other on EmOC, asked of the subset of facilities that did deliveries. In both cases, staff were asked whether the facility could provide a given service or perform a given function. This method of asking about theoretical capability overestimates actual functioning [17, 35], so adding more specific questions, such as whether a given function has been performed within the last 3 or 6 months, as if often done in EmOC assessments, reduces estimates of capability. We did not have information on actual performance in a recent recall period, but we added criteria on staffing, opening hours, communication tools and referral capability to our classification, partly to make our assessment more stringent. We used a similar approach previously for delivery and antenatal care [23, 36]. Of the 1421 facilities in the HFC dataset, 1369 (96 %) had data on both FP and delivery care provision. Of these, 1130 facilities (one of the 1131 facilities we reported on previously was missing data on provision of family planning) offered delivery care. These facilities were asked their capability to provide the eight EmOC signal functions including manual removal of retained products, parenteral antibiotics, parenteral oxytocics, caesarean section (surgical capability), and blood transfusion among others. The 239 facilities that did not provide delivery care were asked whether they provided PAC, but not about EmOC signal functions. Twenty-two had missing data for this variable. We used data from the facilities that provided delivery care (and thus had information on both the general PAC question and the specific signal functions) to assess the validity of using the PAC question alone. There were 834 delivery facilities (74 % of 1130) that reported doing PAC, of which 69 % (578) stated they could remove retained products, 81 % (673) could provide parenteral antibiotics, 67 % (557) could provide parenteral oxytocics, and 43 % (360) could do all three functions. Stratified by facility level, 94 % (81) of hospitals that said they provided PAC also said they provided all three functions, compared to 37 % (267) among health centres and 39 % (12) among health posts. Nevertheless, despite the low validity of the PAC question at lower levels, we assumed that facilities that claimed to provide PAC services actually did so, if we did not have information on signal functions. To be consistent and not disadvantage facilities with more information available, we classified facilities offering all three PAC signal functions as able to provide PAC even if they answered “No” to the PAC question (n = 49). Other missing responses for the three signal function variables (<1 % among delivery facilities) were coded as “no”, in other words, as an inability to provide a particular intervention. However, for blood transfusion and surgical capability (measured via caesarean section), which are both required for comprehensive PAC, we assumed facilities with missing information were unable to provide these functions, including for non-delivery facilities where this information was not asked. The question on 24/7 staffing was also only asked of delivery facilities and was thus missing in 56 non-delivery facilities that reported providing PAC. While such facilities were unlikely to provide 24/7 PAC services, we nevertheless assumed they had such opening times. Where staffing information was entirely missing (in 160 facilities, 12 %), we assumed the required staff members were not available. To categorize services in Zambia, we allowed for three scenarios: All signal functions required, including FP; three doctors in the facility required for legal TOP procedures in non-emergencies. Family planning (as in Table 1) was required because it is best practice for abortion service provision [2, 3]. Three doctors were required because Zambia’s abortion law permits pregnancy termination on a wide range of health and socioeconomic grounds but requires in a non-emergency that abortion be performed with the consent of three registered medical practitioners, one of whom must be a specialist with expertise relating to the case. In emergency situations, consent from only one physician is needed [37], though some providers assume that the risks posed by unsafe abortion mean all cases are emergencies [38]. Because Zambian law makes no specification for PAC [38], this scenario was only applied to TOP. All signal functions required, including FP; three doctors were not needed for TOP, rather one mid-level health professional was adequate, because mid-level providers have been shown effective for this task [39–42]. Family planning was required because this is best practice for abortion service provision. All signal functions except FP required; as in best-practice scenario, only one mid-level health professional (not three doctors) was needed for TOP. This minimal approach focused on the provision of the abortion itself or management of complications, without ensuring an integrated service. We used a range of analytical measures to report on facility capability, estimate national and ward-level populations, and map facilities in relation to the population distribution. To describe facility capability, we used frequencies and percentages. Missing data were handled as described above. The decennial 2000 Zambian Census of Population and Housing [43] contains population numbers down to the ward level, with geographic data on administrative boundaries (provinces, districts, constituencies, and wards) and population growth rates by district. We used the district-level population growth rates to compute ward populations in 2005. To compute the total Zambian population in 2005 for Table 4, we used the national rate of population growth of 2.85 % per year between 2000 and 2010, obtained from the decennial 2010 Zambian Census of Population and Housing. Benchmarks according to Safe Abortion Care (SAC) criteria proposed by Healy and colleagues [8] aAccording to Table 1 criteria We mapped health facilities and ward areas in the geographic information system platform ArcGIS 9.2 (Esri, Redlands, California, USA) and created circles of 15 km radius around each TOP or PAC facility to calculate the proportion of total area covered and the proportion of the ward population within 15 km of services. Lacking higher resolution data, we had to assume an even spatial population distribution inside wards. In rural Zambia, motorized transport is scarce [44, 45] and around 2005, only 1 % of households owned any [46], which means women often had to walk. We used 15 km distance from services to conform to the UN benchmark of 3 hours of travel time [20], assuming a walking speed of 5 km per hour. Geo-location data (Global Positioning System (GPS) coordinates) were available for 1344 of the 1369 facilities; most of those with missing geographic coordinates were military facilities. Ethical approval for the secondary data analysis was granted by the London School of Hygiene & Tropical Medicine ethics committee on 03 July 2007 (application number 5172).

N/A

• Dima Health