Practice Patterns and Perceived Barriers to Dementia Care
Practice Patterns and Perceived Barriers to Dementia Care
An initial draft questionnaire consisting of approximately 50 items was developed in an iterative fashion by the study investigators and consultants in the field of geriatrics. This draft survey then was sent to selected physician members of the AAFP NRN who had agreed to review study materials. Sixteen members reviewed the survey, after which the original questionnaire was reduced to 29 items and approved by the AAFP Institutional Review Board in March 2008.
We selected a random sample from the AAFP's master membership database (N = 60,129) for December 31, 2007. AAFP physicians who spent at least 50% time in direct patient care and reported a mailing address within the 50 United States or the District of Columbia were considered for the survey. From this eligible population of 30,174 members, we selected a random sample of 1,500 using a computerized random sampling strategy. A final sample was randomly selected from a cluster of members based on the randomly generated 7-digit AAFP identifier given to each member. We determined the number of 1,500 was large enough by conventional standards to adequately describe this population within an acceptable margin of error for both 5-point Likert items (95% ± 0.12%; estimated variance of 0.20), and dichotomous survey items (95% ± 5.0%; estimated P = .50), both assuming a 50% response rate, which was exceeded (final response rate, 60%). In June 2008 we mailed the final survey, along with a $2.00 bill, to the sampled physicians. Two follow-up mailings, each approximately 30 days after the preceding mailing, were sent to nonrespondents. The $2.00 incentive also was included in the second follow-up mailing.
We designed the survey to investigate the following components of the care process for patients with dementia in primary care settings: screening; diagnostic evaluation; referral; perceived barriers associated with the treatment and management of dementia; dementia-related behaviors, complications, and comorbidities; family and caregiver problems; and the perceived need for better tools to diagnose/treat dementia (survey instrument is available on request). Some aspects of this multifaceted process were represented by a single dichotomous (yes/no) item ("Do you screen?"), whereas others included multiple-response items in which each response was treated as dichotomous (yes/no) and/or treated as a multiple-count item (eg, "perceived barriers to diagnosing and caring for patients with dementia," "challenging family/caregiver problems associated with dementia").
We considered 3 outcomes for this report: (1) physician ordering of brain imaging as part of screening and/or diagnostic evaluation for dementia (yes/no), (2) assessment of patients who screened positive using more sensitive tests (yes/no), and (3) referral of patients with suspected dementia to other specialists for treatment (yes/no). Other dichotomous outcome items were not selected for further analysis or reporting because of their relative lack of variability. Full results are available on request. The 9 independent variables from the survey instrument were physician respondent's age group and sex; percentage of patient population ≥65 years old; geographic location of practice (urban, suburban, rural); perceived comorbidities, behaviors/complications, family/caregiver problems associated with dementia; number of tools needed to better diagnose and treat dementia; and the number of perceived practice-related barriers to diagnosing and treating patients with dementia. We also included 2 items from the AAFP Master Membership Database: years since medical school graduation and percentage of time spent in direct patient care.
This study was not designed for hypothesis testing, and therefore most analyses conducted were descriptive. Preliminary analysis included univariate and bivariate techniques for the 3 outcome items and independent variables. We used χ, Pearson correlation, and t tests to assess possible bivariate associations between the named outcomes and each of the independent variables. Those items found to be statistically associated with a given dependent variable (P ≤ .20) then were entered simultaneously into a multivariate logistic regression analysis for each outcome with the other items excluded. We used SPSS software version 17.0.1 (IBM/SPSS Inc., Chicago, IL) and SAS software version 9.3 (SAS Inc., Cary, NC) for the analyses and included summary descriptions of demographic data and frequency data of survey answers.
Methods
An initial draft questionnaire consisting of approximately 50 items was developed in an iterative fashion by the study investigators and consultants in the field of geriatrics. This draft survey then was sent to selected physician members of the AAFP NRN who had agreed to review study materials. Sixteen members reviewed the survey, after which the original questionnaire was reduced to 29 items and approved by the AAFP Institutional Review Board in March 2008.
We selected a random sample from the AAFP's master membership database (N = 60,129) for December 31, 2007. AAFP physicians who spent at least 50% time in direct patient care and reported a mailing address within the 50 United States or the District of Columbia were considered for the survey. From this eligible population of 30,174 members, we selected a random sample of 1,500 using a computerized random sampling strategy. A final sample was randomly selected from a cluster of members based on the randomly generated 7-digit AAFP identifier given to each member. We determined the number of 1,500 was large enough by conventional standards to adequately describe this population within an acceptable margin of error for both 5-point Likert items (95% ± 0.12%; estimated variance of 0.20), and dichotomous survey items (95% ± 5.0%; estimated P = .50), both assuming a 50% response rate, which was exceeded (final response rate, 60%). In June 2008 we mailed the final survey, along with a $2.00 bill, to the sampled physicians. Two follow-up mailings, each approximately 30 days after the preceding mailing, were sent to nonrespondents. The $2.00 incentive also was included in the second follow-up mailing.
Survey Items
We designed the survey to investigate the following components of the care process for patients with dementia in primary care settings: screening; diagnostic evaluation; referral; perceived barriers associated with the treatment and management of dementia; dementia-related behaviors, complications, and comorbidities; family and caregiver problems; and the perceived need for better tools to diagnose/treat dementia (survey instrument is available on request). Some aspects of this multifaceted process were represented by a single dichotomous (yes/no) item ("Do you screen?"), whereas others included multiple-response items in which each response was treated as dichotomous (yes/no) and/or treated as a multiple-count item (eg, "perceived barriers to diagnosing and caring for patients with dementia," "challenging family/caregiver problems associated with dementia").
We considered 3 outcomes for this report: (1) physician ordering of brain imaging as part of screening and/or diagnostic evaluation for dementia (yes/no), (2) assessment of patients who screened positive using more sensitive tests (yes/no), and (3) referral of patients with suspected dementia to other specialists for treatment (yes/no). Other dichotomous outcome items were not selected for further analysis or reporting because of their relative lack of variability. Full results are available on request. The 9 independent variables from the survey instrument were physician respondent's age group and sex; percentage of patient population ≥65 years old; geographic location of practice (urban, suburban, rural); perceived comorbidities, behaviors/complications, family/caregiver problems associated with dementia; number of tools needed to better diagnose and treat dementia; and the number of perceived practice-related barriers to diagnosing and treating patients with dementia. We also included 2 items from the AAFP Master Membership Database: years since medical school graduation and percentage of time spent in direct patient care.
Statistical Analysis
This study was not designed for hypothesis testing, and therefore most analyses conducted were descriptive. Preliminary analysis included univariate and bivariate techniques for the 3 outcome items and independent variables. We used χ, Pearson correlation, and t tests to assess possible bivariate associations between the named outcomes and each of the independent variables. Those items found to be statistically associated with a given dependent variable (P ≤ .20) then were entered simultaneously into a multivariate logistic regression analysis for each outcome with the other items excluded. We used SPSS software version 17.0.1 (IBM/SPSS Inc., Chicago, IL) and SAS software version 9.3 (SAS Inc., Cary, NC) for the analyses and included summary descriptions of demographic data and frequency data of survey answers.
Source...