The study was conducted in two countries with mature HIV/AIDS epidemics, Cambodia and the Dominican Republic. Data were collected as part of a larger methodological study of HIV/AIDS care and support measures.
Sampling and study methods in Dominican Republic
Three peri-urban areas were studied: Santo Domingo, La Romana, and Puerto Plata. The national prevalence of HIV in Dominican Republic was 0.9 percent for person 15–49 years of age in 2002.
Households were selected into the sample from a list of members of networks of people living with HIV/AIDS (PLWHA). Six established networks of PLWHA in the three peri-urban areas participated in the study. These networks were developed in poor and marginalized neighbourhoods and populations. Persons aged 18 to 59 years living in the identified network households were selected. All persons in the household who self identifies as HIV+ or were chronically ill were interviewed. Those who were chronically ill but had not disclosed their HIV status may have been HIV+ but were not known to network organizers. The questionnaire did not document HIV status. In cases of persons too sick to respond, care givers in the households acted as proxy respondents.
The survey team was trained over a one week period to administer the tool, to collect sensitive information, and to give education on HIV/AIDS. The final team included six interviewers, three supervisors, and four PLWHA guides who took interviewers to the households.
Sampling and study methods Cambodia
Two Cambodia district were studied: rural Maung Russei district of Battambang Province and Siem Reap city. Cambodia experiences a generalized HIV/AIDS epidemic with 2.6% of person 15–59 positive. The district studied were of the same approximate level of HIV prevalence compared to the national average
In Cambodia two distinct sampling strategies were used. In Maung Russei, the survey questionnaires were administered to the head of households with people living with HIV/AIDS (PLWHA). Villages to visit were selected randomly from a list of village identified through two PLWHA support groups. All households including members of the PLWHA support group in selected villages were included. In Siem Riep, respondents were recruited from outpatients at the Chronic Disease Clinic at the referral hospital. All exiting patients 18 to 59 years of age were asked to participate in the study.
No refusal to participate was recorded in either the Dominican Republic or Cambodia. This is thought to be true because the lists of potential households were constructed by network leaders who were knowledgeable about the households and listed only those households that were likely to participate. The gift of a bag of groceries also created an incentive for participation in the study. Item non-response was negligible.
The study instruments included a consent form, a household roster of usual family members, age and sex, chronic illness status of persons aged 18–59. Chronic illness was defined as the inability to function or perform one's normal life role and activities due to illness and/or bed-ridden for three or more months (not necessarily consecutive) in the past year.
The questionnaire items relating to pain and other symptoms were adapted from the Palliative Outcomes Scale (POS). These items asked "In the past 30 days have you been affected by pain": responses were not at all, somewhat, moderate, severe, overwhelming. For those who experienced pain respondents were further asked "was the pain controlled?": no, a little, most of the time, all of the time. The same questions were then asked regarding occurrence and control of symptoms, with examples given of nausea, coughing, diarrhea, and constipation.
Other questions inquiring into the patients' greatest need related to the illness and services received were evaluated for their feasibility for use in population based surveys. [16, 17] Questionnaire items were devised to address access to external support, defined as support given from formal organizations/institutions rather than family, friends, or neighbours, services for medical care, psychological and emotional support, and other social support services.
The questionnaire and instructions were translated into Spanish and Khmer. Backwards and forwards translations were conducted. The translation teams adjusted differences by choosing appropriate wording.
This paper assesses the construct validity of survey questionnaire items concerning pain, pain control, symptom and symptom control. Construct validity addresses the relationship between operational measurement and theoretical concepts. Our clinical experience leads us to expect that chronically ill persons will report more pain and more symptoms than not chronically persons. We also believe that control of pain should not vary by chronic disease status. That is, both the chronically ill and the not chronically ill should report the same degree of pain control at each level of reported pain. If the data are consistent with these expectations we will argue that this supports the construct validity of these questions. This should be true in settings in which pain control for the acutely or chronically ill are not available. The consistencies in the data and the agreement with observed patterns in the data that are predicted by theories are used to support construct validity.
An analysis was conducted comparing measures of reported chronic illness with measures of reported pain and symptoms. A Chi square test was used to test the difference in levels of pain and symptoms between chronically ill groups and non-chronically ill groups. The selection process of subjects into the study was thought to be independent of relationships between the variables studied which is an assumption of Chi square hypothesis testing.
Mean weighted scores for pain at levels of pain control were calculated by assigning each level of pain a score. A score of one was assigned to the lowest level of pain (no pain) and a score of five to the highest level (overwhelming). A similar score system was used for symptoms.
A Mantel Haenszel test was performed to evaluate if the relationship between pain control and chronically ill status was confounded by the level of reported pain. This summary Chi-square test for stratified data (chronically ill and non chronically ill) in this study was used evaluate reported pain control and symptoms control (both nominal variables) for confounding with level of reported pain and symptoms.
The analytical design of this study did not include HIV status of individuals. That is, the questionnaire did not ask about the HIV status of study participants. This study of measures of pain and symptoms sought to validate those measures in a general sense but also examine feasibility of such a study in high HIV/AIDS prevalence areas. The fact the sample persons were recruited from a network of households in which HIV positive persons were known to live according to network organizers is considered as a background characteristic of the study. The high level of HIV+ households in this study address questions about the feasibility of such a study in surveys in high prevalence settings. The data were analyzed using SAS.