A cross sectional survey was undertaken over a 12 month period between November 2009 to October 2010 across a network of 56 public community health facilities in one health district in New South Wales (NSW), Australia. The district provides healthcare to approximately 840,000 people residing in urban, regional, rural and remote locations, and employs over 1,400 clinicians.
The data were obtained prior to the implementation of an intervention trial reported previously , and approved by the Hunter New England Local Health District (No. 09/06/17/4.03) and the University of Newcastle Human Research Ethics Committees (No. H-2010-1116). The trial involved the sequential rollout of an intervention to three geographical and administratively separate groupings of the 56 community health facilities (Group 1 rural and regional, Group 2 regional, rural and remote, Group 3 urban and rural). Facilities in all three groupings involved a similar mix of community health services (nursing, allied health, child and family, diabetes, aged care, and other service types e.g. rehabilitation, chronic and complex care, women's services, migrant services, renal/dialysis, and regional health service programs), with common policies, standards, governance and performance monitoring processes. Services ineligible for inclusion were: sexual assault, palliative care, aged care assessment teams, dementia, home modification, genetics, and child protection services. Such services were deemed ineligible based upon the advice from the clinical services.
Adult clients were eligible to participate if they: had at least one face to face contact with an eligible service within the prior two weeks; spoke English; were mentally and physically capable of completing the interview (determined at interview, or by clinician/family discretion prior); and were not involved in another community health study.
Each week, over 12 consecutive months, approximately 48 clients from each of the three facility groupings  within the health district were randomly selected from client electronic medical records. Selected clients were mailed an information letter and telephoned to determine eligibility and consent. Data were obtained via computer-assisted telephone interviews (CATI) (approximately 25 minutes), and from client electronic medical records.
Client and service characteristic information collected by the CATI included: employment status; Aboriginal or Torres Strait Islander status; marital status; highest level of education achieved; and conditions in the prior two months for which the client needed to take medication or receive medical attention. Client age, gender, country of birth, postcode, service attended (nursing, allied health, child and family, diabetes, aged care, and other service types) and number of visits to the service in the prior 12 months were obtained from client medical records.
Clients were asked to indicate, in the month before seeing the service: their frequency of smoking tobacco products ; the number of serves of fruit and of vegetables typically eaten per day ; how often they had a drink containing alcohol, the number of standard drinks consumed on a typical drinking day, and how often they consumed four or more standard drinks on any one occasion ; and how many days a week they usually did 30 minutes or more of physical activity .
Based on national guidelines [53–56], clients were considered to be ‘at risk’ and hence require a preventive health education response if they reported that they: smoked any tobacco products ; ate less than two serves of fruit or five serves of vegetables per day ; drank more than two standard alcoholic drinks on a typical drinking day or four or more standard drinks on any one occasion ; or engaged in less than 30 minutes of physical activity on at least five days of the week .
The prevalence of three forms of preventive care was measured. For assessment of behavioral risk status, clients were asked if, during an appointment with the service, the clinician asked: if they smoked any tobacco products; how much fruit and how many vegetables they ate; how much alcohol they drank; and how much physical activity they participated in (yes, no, don’t know).
For provision of brief advice, clients who reported being ‘at risk’ as defined by national guidelines [53–56] were asked whether the clinician advised them: to quit smoking or consider using Nicotine Replacement Therapy; to eat more fruit and/or more vegetables; to reduce the amount of alcohol they consume; or to do more physical activity (yes, no, don’t know).
For the provision of referral/follow-up care, clients who reported smoking were asked if they were offered referral to a free NSW Quitline telephone service. Clients who reported inadequate fruit and/or vegetable intake or physical inactivity were asked if they were offered referral to a free ‘Get Healthy Information and Coaching’ telephone service. Clients with ‘at risk’ alcohol use were asked if they were advised to visit their General Practitioner/Aboriginal Medical Service (GP/AMS). All ‘at risk’ clients were asked if the clinician offered to send a summary of their health risks to their GP/AMS (yes, no, don’t know).
To measure client acceptability of preventive care provision, all clients were asked if it was acceptable for clinicians to assess their health risk behaviors, and for ‘at risk’ clients, if it was acceptable for clinicians to provide brief advice and arrange further support for each risk individually and for all risks combined (strongly disagree, disagree, unsure, agree, strongly agree).
Statistical analyses were undertaken using SAS (version 9.2). Client residential postcodes were used to determine disadvantage (Socio-Economic Indexes for Areas [SEIFA]; cut points: higher NSW half [>991] versus lower NSW half [<=991])  and remoteness (Access/Remoteness Index of Australia [ARIA]; major cities versus regional remote towns) . Descriptive statistics were used to describe client and service characteristics. Comparison of participant and non-participant characteristics was undertaken using chi-square analyses (p < .01).
Descriptive statistics were used to examine for each risk separately and for all risks combined, the prevalence of: clients who were ‘at risk’; clinician assessment of risk; clinician brief advice; clinician offer of referral/follow-up; and client acceptability of such care. The analyses were weighted based on the three facility groupings  to ensure that the sample was representative of all clients attending the community health services across the health district. The prevalence of care provision for all risks combined was defined as: assessment for all four risks; the provision of brief advice for all of a client’s self-reported risks; and either an offer to send a health risk summary of all their risks to the clients GP/AMS, or an offer of referral/follow-up for all their risks individually (i.e. telephone counseling, or GP/AMS for alcohol).
Associations between client/service characteristics and the provision of each form of care (assessment, brief advice and an offer of referral/follow-up for each individual risk and for all risks combined) were initially analyzed using chi-square analysis. Logistic regression analyses were subsequently undertaken separately for the provision of assessment and brief advice for each of the four individual risks (8 regression models). Regression analyses were not undertaken for referral/follow-up for the four individual risks due to inadequate sample size. Separate chi-square and regression analyses were similarly undertaken to determine such associations with the provision of all three forms of care (assessment, brief advice and referral/follow-up) for all risks combined (3 models). For each of the regression models the analysis was adjusted for the three facility groupings to account for potential cluster effect. Variables with a p-value of 0.20 or less from the chi-square analyses were included in each separate regression model, utilizing a backward stepwise selection process whereby the variable with the highest p value was removed until all predictors in the model had a p value less than .01. Any potential interaction between variables that remained in each of the final models were also examined to ensure the model was sound and that results for each variable could be interpreted independently from other variables .