NOTE: "States" in this document includes the District of Columbia, Guam, Puerto Rico, and the Virgin Islands.
Policy Memo 2001.1 is effective starting with the 2002 data year. It replaces Policy Memos 2000.3 and 2000.4 and the "Interpretations of BRFSS Disposition Rules" document produced by the BRFSS Data Collection Methods Committee.
This document presents a minimum set of final and interim codes for use with the BRFSS survey. States may use more detailed interim and final disposition codes internally as long as the interim disposition codes used are subsets of those listed below and the interim and final disposition codes sent to the Behavioral Surveillance Branch (BSB) are those listed below. The one, major, exception is that states using the WinCATI questionnaire supported by BSB can use only the interim and final codes specified below. States may also treat the callback rules as minimal standards and adopt more strenuous efforts to complete interviews.
Section II discusses the criteria used to decide on final and interim disposition codes. Section III describes the process for assigning final disposition codes that require taking into account the outcomes from more than one attempt. Section IV presents the definitions of and callback rules for final disposition codes. Section V presents the definitions of and callback rules for interim disposition codes. Finally, Section VI describes the rules for assigning particular final disposition codes from specific patterns of interim disposition codes.
Note: The term "respondent" used without a preceding adjective of "selected" refers to anyone who answers the telephone.
Final disposition codes serve at least four purposes for four usually distinct groups of people. All four purposes relate to data quality but from different perspectives.
From a research perspective, the major concern is the potential for bias in the data. Researchers are most interested in response rates, which, conceptually, are the number of completes divided by an estimate of the number of eligible households in the sample. From this perspective, telephone numbers for which it is unknown whether or not they ring into an eligible household should be assigned disposition codes that reflect as much information as is known about the telephone number. These codes should distinguish as much as possible telephone numbers that have a large probability of ringing into an eligible household from those that have a small probability of ringing into an eligible household.
From a sponsor’s perspective, the major concern is adherence to protocol. (A sponsor in this context is any organization with a financial stake in the BRFSS survey.) A sponsor wants to know how good a job the data collector is doing. Even conceptually, however, response rates should be affected by the characteristics of the target population in addition to the performance of the data collector. (And in practice, response rates are also affected by sample design and characteristics of the telephone system.) But the data collector has control only over its performance. Protocol defines the standards that the data collector should meet and those standards have been designed to produce acceptable quality data. From this perspective, disposition codes should distinguish between telephone numbers to which different protocols apply. For example, the difference between a refusal from a selected respondent and a postponement from a selected respondent that never results in a completed interview is important from this perspective because different callback rules apply. This same difference, however, is unimportant from a potential-for-bias perspective because they are both incompletes from an eligible household.
From a data collection perspective, the major concern is operational control. Although adherence to protocol is part of operational control, efficiency and interviewer performance are more important than from other perspectives. For example, the distinction between "Language problem after respondent selection" and "Physical or mental impairment after respondent selection" is unimportant from other perspectives but is important from an operational perspective.
From a weighting perspective, the major concern is to identify the amount of nonresponse that occurs at each point at which nonresponse is possible. The collection of information on the point in the interview process at which a nonresponse occurs allows explicit adjustment for nonresponse at each point. This, in turn, should lead to more accurate estimates of statistical parameters, such as means and regression coefficients. From a sponsor’s perspective, how much nonresponse occurs each point at which it is possible is less important than the total amount of nonresponse.
The need to satisfy at least four distinct purposes that imply different criteria for the distinctions among final disposition codes, has resulted in the adoption of thirty-one final disposition codes.
When the BRFSS first began in 1984, most states conducted the survey on paper and there were only eleven final disposition codes. In such a situation, it was reasonable to expect the interviewer to be aware of the call history of each piece of sample and to assign a final disposition accordingly. With the adoption of CATI packages in all BRFSS states and advances in the capabilities of CATI packages, a different model for assigning disposition codes is now possible. For many years, DOS Ci3 CATI and now WinCATI have assigned final disposition codes to records with fifteen or more attempts. Over the past few years, BRFSS has been moving even further towards a model where the interviewer assigns a disposition code (which may be interim or final) after each attempt strictly based on what happened on that attempt only. For example, in data year 2001, BRFSS protocol for the first time specified the assignment of final disposition codes based on patterns of previously assigned interim disposition codes. It was envisioned that at least some states would program the assignment of final disposition codes and that interviewers would be assigning final disposition codes only in cases where the outcome of a single call attempt dictated a final disposition, for example a complete or a non-working number.
In 2002, we are taking further steps towards the adoption of a model where the interviewer assigns a disposition code (which may be interim or final) after each attempt based strictly on what happened on that attempt only. If an interim disposition code is assigned, then the CATI package reviews all the interim disposition codes for that telephone number to determine if a final disposition code is warranted. For example, if a respondent (non-irately) refuses to continue, the interviewer would code the attempt as an interim refusal. The CATI package then looks to see if the refusal was a second refusal and, if so, at what point during the interview process the refusal was made. It then assigns the appropriate final disposition code and retires the number. This process is run after each attempt that is given an interim disposition.
There are at least three advantages to such an approach: (1) The interviewer needs to remember a smaller number of interim and final disposition codes than would otherwise be the case. (2) The interviewer assigns only those final disposition codes that depend on what happened in a single attempt. Thus, the interviewer does not need to be aware of the call history of a number to assign a final disposition code. (Of course, the interviewer may still need to be aware of the call history of a number in order to try to complete an interview.) (3) Human errors in the assignment of final disposition codes based on interim disposition codes are eliminated (assuming correct programming). The disadvantages are that it will probably require additional programming to implement the approach and that not all CATI packages may be capable of implementing it.
Beginning with data year 2002, WinCATI will be programmed to implement this approach. Please note that WinCATI users of the BSB-prepared questionnaire will be able to call a number more than 15 times only by scheduling an appointment on the fifteenth and each subsequent call attempt.
The interim disposition codes are the minimal set implied by the final disposition codes and two additional codes — "Null attempt" and "Requires supervisor attention" — that are useful for operational purposes.
120 Partial Complete
2. Non-Interview, Household With Eligible Respondent
210 Termination within questionnaire
220 Refusal after respondent selection
230 Selected respondent never reached or was reached but did not begin
interview during interviewing period
240 Selected respondent away from residence during the entire
250 Language problem after respondent selection
260 Selected respondent physically or mentally unable to complete an
interview during the entire interviewing period
270 Hang up or termination after number of adults recorded but before
280 Household contact after number of adults recorded but before
3. Non-Interview, Eligibility Undetermined
305 Household members away from residence during entire interviewing
310 Hang-up or termination, housing unit, unknown if eligible
315 Household contact, eligibility undetermined
320 Language problem before respondent selection
325 Physical or mental impairment before respondent selection
330 Hang-up or termination, unknown if private residence
332 Contact, unknown if private residence
335 Telephone answering device, message confirms private residential
340 Telecommunication technological barrier, message confirms private
345 Telephone answering device, not sure if private residence
350 Telecommunication technological barrier, not sure if private
355 Telephone number has changed status from household or possible
household to non-working during the interviewing period
360 No answer
370 On never call list
4. Not Eligible
410 Household, no eligible respondent
420 Not a private residence
430 Dedicated fax/data/modem line with no human contact
440 Fast busy
450 Non-working/disconnected number
505 Refusal: hang-up or termination
515 Language problem
520 Physical or mental impairment
525 Answering machine, message confirms residential status
530 Technological barrier other than answering machine, message
confirms residential status
535 Answering machine, not sure if private residence
540 Technological barrier other than answering machine, not sure if
545 Phone number temporarily out of service
550 No answer
565 Fast busy
570 Possible non-working number
575 Circuit busy
580 Null attempt
585 Requires supervisor attention
In the process described earlier, after each call attempt where the interviewer has assigned an interim disposition code, the CATI package reviews all the interim disposition codes that the telephone number has received and, if appropriate, assigns a final disposition code to that telephone number. The table below shows the appropriate final disposition that should be assigned in such cases. States should contact their Project Officer about cases not covered by the table below.
States should ensure that every record whose last disposition in its call history is an interim disposition is assigned a final disposition according to the following rules. The rules below are hierarchical. The first rule should be applied first to all records with a last, interim disposition code, then the second rule to the remaining records, etc. As noted above, beginning with data year 2002, WinCATI will be programmed to implement this approach.
A final disposition code of 450 Non-working/disconnected number may need to be converted to a final code of 355 Telephone number has changed status from household or possible household to non-working during the interviewing period. Note that this conversion is programmed into the Ci3 questionnaire in WinCATI. States using other CATI system should configure their CATI systems to make the change or they should make the change during their post-data collection processing.
When an interviewer assigns a final disposition code of 450
Non-working/disconnected number on a second or subsequent call attempt, the
previous disposition codes should be reviewed. If at least one previous code
is 505, 510, 515, 520, 525, 530, 535, or 540, then the final disposition
code should be changed to 355 Telephone number has changed status from
household or possible household to non-working during the interviewing
period. This situation is included in the table below even though it
involves changing a final (as opposed to interim) disposition code.
This page last reviewed April 20, 2012[00_pgfooter2.htm]