Data Collection and Protocol
Two prospective open-cohort studies were conducted in Oshikhandass village, from September 16, 1989 to September 15, 1996 (Study 1) and November 11, 2011 to March 31, 2014 (Study 2); surveillance in Study 2 was interrupted from early February until late March, 2012 due to political unrest. All children in Oshikhandass under age 60 months with parental consent were enrolled, including those born and migrating into the village, and followed to 60 months, out-migration, or death; the total number of children was updated continually throughout the study. Women from the village, nominated by local women’s organizations, were recruited and trained in World Health Organization (WHO) guidelines. These women visited children weekly and asked mothers whether the child had diarrhea in the past week. From December 10, 1992 to June 18, 1996 (Study 1) and January 30, 2012 to March 31, 2014 (Study 2), pneumonia was added to the study protocols and mothers were also asked about cough and difficulty breathing. During these visits study staff provided information to mothers on the causes and management of diarrhea and pneumonia and promoted breastfeeding (including exclusive breastfeeding up to 4 months in Study 1 and 6 months in Study 2), immunization, hygiene, and boiling of drinking water. During Study 2, a survey was conducted to determine mothers’ knowledge of the causes of diarrhea and pneumonia and treatment methods.
Supervision in Study 1 was provided by two LHVs, a doctor, and the principal investigator of the study (ZR); and starting in December 1992, when pneumonia was added to the study protocol, study staff were available daily at the government dispensary to examine and treat children with pneumonia between study visits. In Study 2 supervision was provided by a Senior LHV and two sociologists, and study staff were not available in a central location on days outside of regular home visits. In both studies, children with at least one visit after enrollment were included for analysis. Children who died before they were visited were included for mortality, but not morbidity analyses. If a child died anytime while under follow-up, the date and cause of death were recorded from verbal autopsies.
Diarrhea was defined as having three or more loose or liquid stools per day or bloody diarrhea (dysentery), by maternal report, at any time. Diarrheal episodes were defined as consecutive days of diarrhea separated by ≥2 diarrhea-free daysnd categorized by duration: acute, <7 days; prolonged, 7 to 13 days; and persistent ≥14 days. In case of diarrhea, history was taken of bloody stools, tenesmus, fever, vomiting frequency, breastfeeding status, other milk consumption, appetite, thirst and medications used; examination included general status (alert, unwell/irritable/lethargic, very sleepy/unconscious), skin turgor, temperature (axillary), rectal prolapse and weight. Study staff provided oral rehydration packets, zinc supplements (Study 2 only), nutritional advice, antibiotics for dysentery (Study 1 cotrimoxazole; Study 2 ciprofloxacin) or referred the child for general danger signs, severe dehydration or malnutrition.
If the mother reported the child had cough and/or difficult breathing, project staff assessed for pneumonia. Pneumonia was diagnosed as per WHO guidelines, however these changed slightly between the two studies. In Study 1, pneumonia was defined as cough with fast breathing based on age-specific cutoff breathing rates using a one-minute Acute Respiratory Infection timer [23], presence of chest indrawing (severe pneumonia), and associated danger signs (very severe disease): unable to drink, convulsions, abnormally sleepy/difficult to wake, stridor in a calm child, severe malnutrition [24]. By Study 2 danger signs had been modified to include also the inability to drink or eat anything, or vomiting everything [25]. Infants <2 months with pneumonia were required to have immediate referral and were excluded from pneumonia rate calculations as data collection was incomplete for this age group.
Children with non-severe pneumonia were treated for 5 days with cotrimoxazole in Study 1, and with 3 days of cotrimoxazole or amoxicillin in Study 2 [26]. Children with severe pneumonia were given an initial dose of amoxicillin and referred immediately to a physician. In Study 2, children with wheezing were given up to three bronchodilator treatments via spacer or nebulizer in the study office prior to diagnosis. All children with pneumonia, regardless of severity, were seen by a supervisor the same or next day to make sure the family followed recommendations and the child was not worse. Any child who was not improving after 48-72 hours was changed to a second-line antibiotic. All children were evaluated after 5 days and again after 14 days to ensure symptoms were resolved.
In 1995, Pakistan’s Lady Health Worker (LHW) Programme [27] was introduced in the community. Of the eight LHWs selected to serve the village, three had been trained as study staff in Study 1. LHWs continued to serve the community during the intervening years between Study 1 and Study 2. Their activities were similar, but less intensive than those of the study staff, with the addition of providing contraceptives to women in the community [27]. Also in the intervening periods between studies, a water filtration plant was built (2002), supplying only part of the community, and H. influenzae type b (Hib) vaccination was initiated by the Expanded Program on Immunization (2009), given as part of a pentavalent vaccine at 6, 10, and 14 weeks of age [28].
In both studies, length/height was measured every three months and weight was measured monthly. Data on socioeconomic status and family characteristics were collected by questionnaire; families were defined as a mother, father and children; households were defined as multiple families which shared the same kitchen. In Study 1, questionnaires were administered initially in September 1989, new families were added as they entered the village, and questionnaires were updated every 6 months to reflect additional family members; in Study 2, family questionnaires were administered initially in November 2011 and updated to include new families until March 2014. In both studies, questionnaires were administered to all households in the village regardless of if a child was enrolled in the study. Where possible comparable categories between Study 1 and Study 2 were used for analysis. Variables included categories of housing based on construction materials (improved: cement, concrete, brick; unimproved: no walls, mud, wood, stone; somewhat improved: a combination of the previously described categories, only used for Study 1); the number of rooms in the house and crowding (people per room). In Study 2, water from the filtration plant constructed in 2002, water brought from Gilgit, and purchased mineral water, were defined as improved drinking water sources. There were no improved drinking water sources in Study 1, but households were asked if they treated their water, which generally consisted of allowing glacier sediment to settle. Improved sanitation facilities included flush toilets (to septic tank or pit latrine) and new composting twin pit latrines; unimproved sanitation facilities included traditional pit latrines and open defecation. Improved cookstoves (Study 2) included gas or electric stoves or Building and Construction Improvement Program stoves with or without water heater [29]; and clean cooking fuels (Study 1) included gas or electricity. Although most households reported using multiple cooking methods, the stove/fuel reported as most frequently used was taken for analysis. In Study 2 information on indoor air pollution was collected through study staff observation and self-report. Information on household income was divided into quintiles and parental education was categorized in 3 levels (category 1: illiterate/ no formal education; category 2: primary – matriculation/Class 10; category 3: intermediate/12 years and above). Mothers were asked if they earned or not and fathers’ occupations were divided into 4 levels based on salary and social status. All data, including childhood health data and socioeconomic variables, were collected on paper forms by study staff and stored securely at the AKHS,P office (Study 1) and Karakorum International University (Study 2). Data were double entered and stored in SPSS (IBM, New York, USA) databases.
Statistical Analysis
Incidence and prevalence are reported in child-years (CY) at risk for diarrhea or pneumonia and described in terms of age and season. Children were considered under follow-up if the interval between visits was ≤15 days; intervals >15 days were excluded from the observation period. Child-years (CY) at risk were calculated by subtracting the number of days a child had diarrhea/pneumonia (excluding the day the episode started) from the total number of days followed and dividing by 365.25. Incidence rates were calculated by dividing the total number of episodes by the total CYs at risk and prevalence calculated by dividing the total days of diarrhea or pneumonia by the total CYs of follow-up. Hurdle models [30] were constructed for each study separately to examine putative risk factors for both diseases, accounting for the large number of children with no reported episodes of diarrhea and pneumonia. The models comprised a logit function for the chances of never having had an episode and a Poisson function to describe the risk of additional episodes of diarrhea or pneumonia with respect to the reference level of the independent variable. A random intercept was included in the model to account for shared facilities between family members [31]. The final hurdle model was used to predict the number of episodes/CY for a unit change for each level of the independent variables. Additionally, simulated data were used to predict the number of episodes/ CY for an idealized scenario (all improved/ highest level categories) and a reference scenario (all unimproved/lowest level variable) for both diarrhea and pneumonia during each study period. Models were constructed with the glmmTMB package [32] in R version 3.5.0 [33].