COST OF UNSAFE ABORTION IN UGANDA / 2010

Manualfor the Principal Investigator and Regional Coordinators

FOR WOMEN’S INITIAL INTERVIEWS, PROVIDER CASE INTERVIEWS AND WOMEN’S FOLLOW-UP INTERVIEWS

Makerere College of Health Sciences/Makerere University, Kampala
Guttmacher Institute, New York
5 January2011
1
COST OF UNSAFE ABORTION IN UGANDA / 2010

I. Sample design—initial interview

The sample of women for interviewing at health facilities using the Women’s Initial Questionnairewill be selected using the following rules:

  1. The women’s initial questionnaire will be administered to around 1790 women undergoing post-abortion care treatment at health facilities across Uganda. The intention is to successfully conduct around 1500 interviews—the larger number includes expected refusals.
  2. The women will be sampled from 39 health facilities across the country.
  3. Among these 39 health facilities are 18 hospitals, 11 regional hospitals, 7 level-four health centers, and 3 level-three health centers.
  4. Twelve of these health facilities are used for regional referral.
  5. Six are government district hospitals. Nine are NGO hospitals. Four are private hospitals.
  6. Eight health facilities are level 4 and below (this includes level 3).
  7. Level 3 health facilities are run by midwives, and have no attending doctor. However since they provide post-abortion care, they are included in the sample.
  8. The bulk of the sample will come from the larger regional hospitals. We will attempt to interview about 1170 women from these hospitals. Another 620 women will be interviewed from the smaller health facilities, which include the level 4 and level 3 health centers, NGO hospitals, private hospitals, and government district hospitals.
  9. About 300 cases will come from MulagoHospital since that is the only national hospital in the sample.
  10. A total of 11 regional coordinators (RIs) will be overseeing the data collection activities. Regional coordinators will be based out of the big regional hospitals, and they will oversee the data collection in the smaller hospitals and health centers in their region.
  11. Interviewers will be based out of either the regional hospitals or the smaller health facilities. Not every health facility will have an interviewer based out of it; but every selected health facility will be covered by an interviewer and a supervisor.

For the low-volume health centers, all women presenting with post-abortion complications will be canvassed. An arrangement with the health providers will be made so that the interviewer will be contacted immediately by phone whenever such a woman presents. The interviewer will make every attempt to interview the woman. If, by any chance, the interviewer is unable to meet the woman, this will be noted as well as any other information available from the health center, including the woman’s age, marital status, number of living children and the diagnosis. A log of all such missed interviews will be kept by the RC supervising the interviewer, who will transmit the log periodically to the principal investigator (PI) who will keep a national log of such occurrences.

For the high-volume hospitals in the sample, a maximum number of interviews by the interviewer will be established by the RC and PI, taking into account the normal duties that the interviewer will still have. The maximum number should not exceed four per day. The following table indicates how interviews should be distributed throughout the day in these cases:

Established maximum no. of interviews per day / Morning hours / Afternoon hours
2 interviews per day / 0-2 interviews / 0-2 interviews
3 interviews per day / 0-2 interviews / 0-2 interviews
4 interviews per day / 0-2 interviews / 0-2 interviews

There should be at least a one-hour break between any two interviews. There should never be more than two interviews during morning hours (8am to 1pm) or afternoon hours (1pm to 6pm). Once these two requirements are met, the interviewer should attempt to interview each new patient who presents with post-abortion complications.

Example 1: A woman presents 40 minutes after the end of an interview. She should not be interviewed, since the time between interviews is less than one hour.

Example 2: A woman presents 90 minutes after the end of an interview. She should be interviewed, since the time between interviews is more than one hour.

Example 3: Two women have been interviewed in the morning and a third woman presents in the afternoon at a hospital where the maximum number of interviews per day has been set at two. The third woman should not be interviewed, since the maximum has been reached.

If any women presenting with post-abortion complications are not interviewed because of these limiting rules, this will be noted as well as any other information available from the hospital, including the woman’s age, marital status, number of living children and the diagnosis. A log of all such missed interviews will be kept by the RC supervising the interview, who will transmit the log periodically to the PI who will keep a national log of such occurrences.

Maximum number of interviews per facility. Limits to the total number of completed interviews per facility have been established ex ante. These are shown in the appendix to this manual. It is important for the survey to adequately sample from different facility levels and from different regions of the country. On the other hand, estimates of the numbers of expected post-abortion cases per facility made ex ante may turn out to be either under or over-estimates. To balance these two concerns, the PI and RCs must maintain an updated status of numbers of interviews at all facilities. In principal, interviewing should be discontinued at facilities once they reach their respective expected numbers of interviews. This is a dynamic process, however, and the PI should make “course correction” decisions as the field work progresses. If some facilities are adding completed interviews at a much slower rate than expected, the maximum number of interviews at other facilities of the same level should be correspondingly raised.

The following table shows the distribution of expected interviews by facility level. To the extent possible, the proportions shown in the table should be maintained, even if some facilities produce fewer interviews than expected and others more than expected.

HC III / HC IV / HOSPITAL / R. HOSP / NAT.HOSP / TOTAL
25 / 90 / 455 / 920 / 300 / 1790
1.4% / 5.0% / 25.4% / 51.4% / 16.8% / 100.0%

II. Sample design—follow-up interview

Refer to Table 1 in the appendix for the required sample sizes by the three categories of respondents listed, namely “Certain/probable induced abortion”, “Possible induced abortion” and “likely miscarriage”. Note that the sampling scheme assumes that the respondents from the initial interview have been classified before the follow-up interviews begin. While 100% of respondents in the first two groups should be followed up, only 50% of those in the third group (miscarriages) should be, according to the schema in Table 1.

Given the one-month gap ideally between the two interviews, it a data entry/data analysis system has been put in place that is capable of selecting the women for the follow-up sample quickly and with that period. This system consists of the following elements:

  1. Interviewers should make every effort to obtain the consent and cooperation of all interviewed women for the follow-up interview since, at the time of the initial interview, it is not known to which group the women belong.
  2. The filled-in questionnaires from initial interviews should be inspected by the RCs, corrected if necessary and reach the PI in Kampala within two weeks.
  3. Filled-in questionnaires from initial interviews should be entered by the Mulago team into an electronic data file within one week of arrival from the field.
  4. A macro for the analysis of the data should be run at least weekly and the list of respondents showing the classification of their abortions (certainly induced, probably induced, possibly induced, or spontaneous) should be sent to the RCs on a weekly basis.
  5. All respondents with certainly, probably or possibly induced abortions should be re-interviewed (unless the woman refuses at this stage). Only 50% of women with miscarriages should be re-interviewed.[1]
  6. To maintain a random method of selection of women with miscarriages, every second woman on the list of miscarriages should be selected for re-interview.

A mid-course correction may be necessary in deciding upon the follow-up sample because the actual proportions of women falling into these three categories may be different from our assumptions. A hypothetical example may make this clearer. Suppose that we are aiming for the following number of follow-up interviews (as in Table 1):

Certain induced abortions260

Possible induced abortions340

Likely miscarriages130

Total730.

Suppose that during the first month of follow-up interviews 250 follow-up interviews take place. At this point, preliminary results show the following:

ActuallyRemain to be

Interviewedinterviewed

Certain induced abortions60200

Possible induced abortions110230

Likely miscarriages8050

Total250480.

One can see in this example that relatively more miscarriages were re-interviewed than desired; hence in the remaining time for the follow-up interviews, the team must be careful to interview relatively few miscarriages (only 50 more in all, in this example). The Mulago team must keep track of what is happening in each region as well as nationally. Instructions should be given to each RC from time to time as to the percentage of respondents with miscarriages that should be re-interviewed. In this example, the overall percentage would fall below 50% and in any particular region the percentage could fall to as low as 0% if that region had already re-interviewed many women with miscarriages.

III. Using a unique number to identify a respondent

The PI will generate a series of unique numbers (with no repetitions) which will be called the “study identification numbers” (SINs). These numbers will be distributed in batches to the RCs who, in turn, will distribute lists of numbers to the interviewers supervised by them. Each time an interviewer begins an interview with a woman, the next available number from her list will be written in the questionnaire (see question Q101 of the Women’s Initial Questionnaire), whether or not the woman is interviewed or refuses to be interviewed. The interviewer must be very careful to cross the number off her list as soon as she has written it onto the questionnaire form to assure that no questionnaires have duplicate numbers.

The interviewer will also write this number along with contact information of the woman in the interviewer’s log. The log will be given to the RC at the end of every week of the survey and a new log started the following week. The information in the log should be:

  1. The woman’s study identification number
  2. The woman’s name
  3. The woman’s home address
  4. The woman’s mobile telephone number
  5. Any particular directions for finding the woman’s home for the follow-up interview.

The number of SINs generated for each facility in the sample should be related to the expected number of completed interviews for that facility. The number generated should be double the number of expected interviews.

The format of survey identification numbers will be: AA/BB/CCC,

Where:“AA” is the number corresponding to the type of facility,

“BB” is the number of a health facility within a particular facility type, and

“CCC” is a consecutive number (say, 001 to 999).

For example, in this format, the SIN 05/12/045 would mean the 45th interview attempted in the 12th health facility of facility type 5.

IV. Procedure to follow when a woman refuses to be interviewed

It is most straightforward and least likely to cause errors to have one questionnaire per respondent, even if that interview is not completed. For instance, at the point of refusal the interviewer would have assigned one of the unique study identification numbers to that questionnaire and crossed it off her list. We would not want to reverse this process, if that interview turns out to be a refusal. Also, this procedure allows for the possibility that the RC will later meet with respondent, if the interviewer gets a refusal, and we want the same questionnaire to be used to record these different steps.Therefore, for each patient contacted there should be a questionnaire with a SIN, no matter what the result of the interview is.

If a woman does not agree to participate in the interview at the health facility, no interview should take place. Since the contact with a potential respondent takes place shortly before the woman is discharged from the facility, there will be no chance to attempt to interview the woman later.

V. Procedure if interviewer cannot speak the woman’s language

Since the interviewers work and function in the areas where they live and work, it should be a rare event for an interviewer to meet a potential respondent who speaks a language which she cannot understand. Therefore, when such cases occur the interview should be terminated. Even though the interviewer and interviewee cannot communicate in such cases, a survey identification number and a questionnaire would have been started. In the interviewer’s note at the end of the questionnaire, the interviewer should describe, as usual, the reason for the non-interview (e.g., “different languages, unable to communicate”) as well as give her impression about the characteristics of the woman.

VI. Number of attempts to contact a respondent for the follow-up interview

The interviewer should make two independent attempts to locate the woman and/or persuade the woman to agree to be re-interviewed. The two attempts should not be on the same day. If, after two such attempts, the woman still will not cooperate, a note should be made in the log describing the attempts and the reasons for refusal, if any.

VII. Field work for the follow-up interviews

The follow-up interviews are supposed to be conducted about one month after the initial interviews, although due to scheduling practicalities they can take place between 30 and 60 days after the initial interviews. Sixty days, however, is the maximum gap permissible since the respondents’ ability to recall events can be expected to be quite weak after two months.

Total number of follow-up interviews. All women who complete the initial interview should be requested to agree to the follow-up interview. We anticipate that a small proportion will refuse at this time. We also expect that another small proportion of women who agreed to the re-interview will change their minds when re-contacted and refuse. Further, another proportion will not be able to be located based on the contact information they provided at the first interview. The loss to follow up cannot be precisely predicted, but the target number of re-interviews should be 700. When the number of completed follow-up questionnaires approaches 600, the PI should make a decision as to whether to stop at 600 (the minimum acceptable) or whether to try for 700. A main factor in this decision should be the state of the budget—remembering that he may request a supplement to the agreed upon budget.

Another factor is the composition of the respondents of the follow-up interview. We would like the composition to be in the following proportions:

  • Women who definitely had an induced abortionat least 35%
  • Women who likely had an induced abortionaround 47%
  • Women who likely had a miscarriageat most 18%.

These percentages are taken from the estimates done for the original proposal for this study (see Table 1 in the appendix). By the time that the field work is getting near to completion, some preliminary analysis should be ready which will help to determine whether the above proportions are being met. In particular, the number of follow-up interviews to women with in the first two categories should be around 500 and the number in the first category should be close to 200. Every effort should be made to reach these numbers. In other words, care must be taken to avoid the third category of women (miscarriages) from being disproportionately high (say higher than 18%).

Simultaneity of the two surveys. The follow-up interviews will overlap with the ongoing initial interviews. The same interviewer who conducted the initial interview should conduct the follow-up interview. After the first month of field work, therefore, she will have to split her time and effort between continuing to do initial interviews at the health facility or facilities to which she has been assigned, and follow-up interviews whose venues will be determined by the respondents, but which in general will require more transportation and logistical effort on the part of the interviewer. The RCs should be involved in the planning and scheduling of the follow-up interviewing process to insure that it proceeds with the least number of problems.

One idea would be for the maximum number of daily interviews in the larger facilities to be revised downward after the follow-up field work has started. For instance, in a facility where the original maximum was four interviews per day, the maximum might be reduced to three initial interviews and one follow-up interview beginning in the second month, with a further revision to two initial and two follow-up visits after three months of field work. This is just an example. The PI and RCs must make such decisions based on a continuous monitoring of progress in data collection based on the factors discussed above.