1. Overview of the SDP surveys

1.1. Data note
country refresh_round note
Burina Faso 5 30 EAs added in R3.
Cote d’Ivoire none -
DRC NONE? 2 EAs dropped in Kinshasa in R2
Ethiopia 5 21 EAs added in R3 (in Oromia).
India, Rajasthan NONE? R2: expanded private SDP sample (up-to 3 SDPs in each contiguous urban EAs and from non-selected rural segments or contiguous rural EAs)
Kenya 3 and 5 31 EAs added in R5 (in Nairobi and two new counties: West Pokot and Kakamega). The number of EAs doubled in the latest, phase-1 survey.
Niger NONE? -
Nigeria, Kano no refresh The number of EAs decreased from 36 to 25 in the latest, phase-1 survey.
Nigeria, Lagos no refresh -
Uganda 5 -
1.2. Number of EAs and SDPs in each survey

The number of EAs and SDPs should be roughly stable, unless there was a change (see above). Note:

  • Burkina Faso Round 6: the number of SDPs decreased without changes in EAs. More details in the next section.
1.3. Distribution of SDPs by managing authority in each survey

The public/private composition also should be roughly stable.

1.4. Distribution of SDPs by type/level in each survey

The type/level composition also should be roughly stable.

2. EA-SDP linking methods

EA-SDP link was done using Stata (see the do file in GitHub). Then a summary dataset describing link results was constructed and used for this report.

Briefly, see the following steps:

A. Collapse HHQFQ dataset to EA-level, including survey round, admin level 1, and urban/rural classification for each EA.

B. Reshape SDP dataset to LONG file, using “EAserved_” variables. Now the dataset has unique EA-SDP pairs. An EA should be paired with multiple SDPs, per survey design. A SDP can be paired with multiple EAs.

  • Technically, all EAs in a survey should appear here, but in some surveys a few EAs are not listed. This is a main (if not the only) reason for EAs not having linked SDPs. This can be caused by various reasons:
  1. The EA truly does not have any SDPs that are “linkable” (unlikely),
  2. It was not possible to survey any of the linked SDPs (due to security reasons, for example), or
  3. Data entry/management errors in “EAserved_” variables in the SDP surveys.

C. Collapse/Create EA-level SDP dataset which has service environment characteristics based on linked SDPs. For example, for each EA, the number of any SDPs offering injectables, the number of primary-level SDPs offering injectables, the number of public SDPs offering injectables, etc.

D. Merge with women datasets (though not necessary for this assessment). If there are missing EAs in above step 2 (and thus 3 too), women in those EAs will be categorized to not have ‘population-level accessibility to quality services’. If the missing EAs are random and minimal, it’s not a big problem.

3. Results

3.1. Number of “EAserved” per SDP, by level of SDPs

3.2. Number of linked SDPs per EA, by type and level of SDPs

Theoretically expected average number of SDPs linked to an EA:

  • Any/all SDPs: between 3-6 (3 public + up to three private), but probably slightly above 3. This includes any kind of SDPs, including pharmacies.
  • Primary or secondary SDPs: between 2-5, but probably slightly above 2. This excludes pharmacies.
  • Public SDPs: 3, possibly 2 depending on health systems (e.g., Ethiopia). This excludes pharmacies, based on design, since there are no public pharmacies.
  • Primary or secondary public SDPs: 2, possibly 1 depending on health systems (e.g., Ethiopia). This excludes pharmacies, based on design, since there are no public pharmacies.

More importantly, the average number should remain stable across all rounds, within a geography. Two potential problems (neither related with EA refresh):

  • Burkina Faso Round 6
  • Niger, Niamey Round 5

  • Nigeria, Kano Round 3

  • average number of SDPs linked to a EA: 3.3 (2 - 5.1)
  • average number of SDPs, excluding hospitals, linked to a EA:2.2 (1.2 - 4)

  • average number of public SDPs linked to a EA: 2.4 (0.8 - 4.9)

3.3. Percent of EAs with no linked SDPs

Theoretically, this should be 0%. But, see the following outliers:

  • Burkina Faso Round 6
  • Niger, Niamey Round 5
  • Nigeria, Kano Round 3

Other than these survey, EA-SDP linking looks ‘okay’ (i.e., <5 %) with any SDPs (dark orange bars) and any public SDPs (dark blue bars). For women in these SDP-missing-EAs, any ‘population-level accessibility to quality services’ will be automatically none/missing.

However, percent of EAs without linked primary or secondary SDPs is high. Median (outliers 54 surveys): 3.7 % for any primary or secondary SDPs (light orange bars) and 10.1 % for public primary or secondary SDPs (light blue bars).

Among 54 surveys:

  • average percent of EAs with no linked SDPs: 2.7 (0 - 38.9). median: 0.2
  • average percent of EAs with no linked SDPs that are not hospitals:6.8 (0 - 56.2). median: 3.7
  • average percent of EAs with no linked public SDPs: 7.8 (0 - 55.6). median: 1.3

Among 51 surveys, excluding the outiers:

  • average percent of EAs with no linked SDPs: 0.9 (0 - 5.8). median: 0
  • average percent of EAs with no linked SDPs that are not hospitals:4.9 (0 - 21.4). median: 3.6
  • average percent of EAs with no linked public SDPs: 5.8 (0 - 39). median: 1.2

3.4. Percent of EAs with no linked SDPs by residential area

Overall, urban EAs tend to have more problems in linking. Excluding four countries/geographies, where urban/rurla stratification was not used because the entire geogrphy is primarily urban/rural.

EA_SDP link code - in Stata

See GitHub for data, code (for both Stata and R), and more information.
For typos, errors, and questions, contact me at yj.choi@isquared.global.

Making Data Delicious, One Byte at a Time