Cancer Prevention
Conditions
Keywords
Cancer prevention, Clinical Decision Support, Rural Health, Shared Decision Making
Brief summary
The objective of this project is to implement and evaluate the effectiveness of a sophisticated Web-based, electronic health record (EHR)-linked clinical decision support (CDS) system designed to improve primary and secondary cancer preventive care. This project will engage a rural population with substantial healthcare disparities and gaps in the receipt of primary and secondary cancer prevention. Results will advance dissemination and implementation research methods that can reduce health disparities and improve healthcare for millions in medically under-served areas.
Detailed description
The objective of this project is to implement and evaluate the effectiveness of a sophisticated Web-based, electronic health record (EHR)-linked clinical decision support (CDS) system designed to improve primary and secondary cancer preventive care. To achieve this objective, EHR data is sent to evidence-based cancer prevention algorithms, housed on a secure Web site. The algorithms: (a) identify at the point of care all eligible patients not up to date on their cancer prevention interventions; and (b) present to both patient and primary care provider (PCP), appropriate evidence-based primary cancer prevention interventions and cancer screening options at the point of care. The cancer prevention (CP)-CDS will focus on breast cancer screening in women aged 50-74, colorectal cancer screening for both sexes aged 50-75, cervical cancer screening for women aged 21-65, human papilloma virus (HPV) vaccination for both sexes aged 11-26, and referrals for weight management and smoking cessation in all adults aged 18 and older. Effectiveness is assessed by cluster-randomizing 30 primary care clinics with roughly 285 PCPs and 153,000 study-eligible patients into one of three experimental conditions: Group 1: CP-CDS intervention in which the decision support is printed for the PCP and patient, designed to engage patients in discussions about appropriate cancer prevention strategies. Group 2: CP-CDS + shared decision making tools (SDMT) intervention in which the CP-CDS is printed along with additional shared decision making tools. Group 3: clinics provide usual care (UC) with no intervention-related activity related to cancer prevention. With 10 clinics, 95 PCPs, and 51,000 potentially eligible patients per study arm, this study will formally test the hypothesis that Groups 1 and 2 are superior to Group 3 over an 12-month follow-up period with respect to: (a) significantly higher rates of appropriate screening for breast, cervix, and colorectal cancer, as defined by the United States Preventive Services Task Force; and (b) significantly higher rates of human papillomavirus (HPV) vaccination in males and females aged 11-26 years. Secondary analysis will evaluate the impact of the intervention on lung cancer screening. The investigators further posit that Groups 1 and 2 will have higher short-term health care costs but better long-term cost-effectiveness than Group 3. The Consolidated Framework for Implementation Research (CFIR) and Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) conceptual frameworks will be used to guide implementation planning, organization, conduct, and impact evaluation of the intervention in a large rural healthcare system. This project will engage a rural population with substantial healthcare disparities and gaps in the receipt of primary and secondary cancer prevention. Results will advance dissemination and implementation research methods that can reduce health disparities and improve healthcare for millions in medically under-served areas.
Interventions
OTHERCP-CDS
The cancer prevention wizard is a point of care clinical decision support system designed to identify patients that are due for preventative cancer services and provides evidenced-based recommendations to the patient and provider.
Sponsors
HealthPartners Institute
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
HEALTH_SERVICES_RESEARCH
Masking
NONE
Eligibility
Sex/Gender
ALL
Age
18 Years to 80 Years
Healthy volunteers
Yes
Inclusion criteria
* meet above age criteria * pcp practice at randomized clinic
Exclusion criteria
* current or past cancer diagnosis (excludes nonmelanoma skin cancer) * hospice care or cancer chemotherapy * Alzheimer's disease codes * major cardiovascular event in 12 months prior to index date
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Composite Cancer Screening Tests up to Date | 1 year post index date | Participants up to date on composite endpoint of breast, cervical, and colorectal cancer screening at 1 year post index date. |
| HPV Vaccination up to Date | 1 year post index date | Participants up to date on HPV vaccination at 1 year post index date. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Lung Cancer Screening up to Date | 1 year post index date | Participants up to date on lung cancer screening at 1 year post index date. |
Countries
United States
Participant flow
Recruitment details
Patients were accrued if they were not up to date on breast, cervical, or colorectal cancer screening, or not up to date on HPV vaccine at the index visit. Eligibility was determined using automated algorithms.
Pre-assignment details
Cluster randomization of the 34 primary care clinics was done to minimize contamination. Pre-randomization balance factors included: 1) clinic urbanicity, and 2) percentage of women at each clinic up to date on breast cancer screening. The first concealed randomization scheme meeting balance criteria was selected. 35953 participants were allocated to one of three groups (i.e. Started). The number Completed reflects the number of subjects included in final analysis after exclusions and opt outs.
Participants by arm
| Arm | Count |
|---|---|
| CP-CDS Providers practicing in the PCP-focused group will receive a point of care clinical decision support tool called cancer prevention wizard (CP Wizard) that is integrated into the EHR. The CP Wizard provides evidenced- based recommendations to the provider and patient related to overdue cancer prevention screening services.
CP-CDS: The cancer prevention wizard is a point of care clinical decision support system designed to identify patients that are due for preventative cancer services and provides evidenced-based recommendations to the patient and provider. | 10,122 |
| CP-CDS + SDMT Providers practicing in the SDMT-focused group will receive a point of care clinical decision support tool called cancer prevention wizard (CP Wizard) that is integrated into the EHR. The CP Wizard provides evidenced- based recommendations to the provider and patient related to overdue cancer prevention screening services. The SDMT-focused arm also provides shared decision making tools to patient and provider at the time of the visit.
CP-CDS: The cancer prevention wizard is a point of care clinical decision support system designed to identify patients that are due for preventative cancer services and provides evidenced-based recommendations to the patient and provider. | 9,034 |
| Usual Care Providers practicing at clinics assigned to the usual care arm will have no access to the CP Wizard. | 12,960 |
| Total | 32,116 |
Baseline characteristics
| Characteristic | CP-CDS | CP-CDS + SDMT | Usual Care | Total |
|---|---|---|---|---|
| Age, Continuous | 48.4 years STANDARD_DEVIATION 17.2 | 47.9 years STANDARD_DEVIATION 17.5 | 47.7 years STANDARD_DEVIATION 17.3 | 48.0 years STANDARD_DEVIATION 17.3 |
| Ethnicity (NIH/OMB) Hispanic or Latino | 116 Participants | 100 Participants | 181 Participants | 397 Participants |
| Ethnicity (NIH/OMB) Not Hispanic or Latino | 9917 Participants | 8863 Participants | 12670 Participants | 31450 Participants |
| Ethnicity (NIH/OMB) Unknown or Not Reported | 89 Participants | 71 Participants | 109 Participants | 269 Participants |
| Race (NIH/OMB) American Indian or Alaska Native | 261 Participants | 169 Participants | 155 Participants | 585 Participants |
| Race (NIH/OMB) Asian | 74 Participants | 65 Participants | 72 Participants | 211 Participants |
| Race (NIH/OMB) Black or African American | 164 Participants | 122 Participants | 199 Participants | 485 Participants |
| Race (NIH/OMB) More than one race | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) Native Hawaiian or Other Pacific Islander | 19 Participants | 9 Participants | 26 Participants | 54 Participants |
| Race (NIH/OMB) Unknown or Not Reported | 145 Participants | 121 Participants | 165 Participants | 431 Participants |
| Race (NIH/OMB) White | 9459 Participants | 8548 Participants | 12343 Participants | 30350 Participants |
| Sex: Female, Male Female | 6725 Participants | 5870 Participants | 8532 Participants | 21127 Participants |
| Sex: Female, Male Male | 3397 Participants | 3164 Participants | 4428 Participants | 10989 Participants |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk | EG002 affected / at risk |
|---|---|---|---|
| deaths Total, all-cause mortality | 22 / 10,122 | 15 / 9,034 | 26 / 12,960 |
| other Total, other adverse events | 0 / 10,122 | 0 / 9,034 | 0 / 12,960 |
| serious Total, serious adverse events | 0 / 10,122 | 0 / 9,034 | 0 / 12,960 |
Outcome results
Primary
Composite Cancer Screening Tests up to Date
Participants up to date on composite endpoint of breast, cervical, and colorectal cancer screening at 1 year post index date.
Time frame: 1 year post index date
Population: Analysis population included adult patients ages 21-75 who had a visit at a study clinic over the accrual period, and who were due for one or more breast, cervical, or colorectal cancer screening test.
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| CP-CDS | Composite Cancer Screening Tests up to Date | 3087 Participants |
| CP-CDS + SDMT | Composite Cancer Screening Tests up to Date | 2590 Participants |
| Usual Care | Composite Cancer Screening Tests up to Date | 3866 Participants |
Comparison: Generalized linear mixed models with logit link and binomial error distribution were used to test the effect of the intervention on the composite of cancer screening by 1 year post-index for breast, cervical, and colorectal cancer. The model included two primary contrasts: CP-CDS vs. UC and CP-CDS+SDMT vs. UC and included pre-specified covariates and variables used in the randomization process. Tests of the two contrasts were 2-sided with P-values \< 0.025 considered statistically significant.p-value: 0.025Mixed Models Analysis
Comparison: Generalized linear mixed models with a logit link and binomial error distribution were used to test the effect of the intervention on the composite of cancer screening by 1 year post-index date for breast, cervical, and colorectal cancer. Model included two primary contrasts: CP-CDS vs. UC and CP-CDS+SDMT vs. UC and included pre-specified covariates and variables used in randomization process. Tests of the planned contrasts were 2-sided with P-values \< 0.025 considered statistically significant.p-value: 0.025Mixed Models Analysis
Primary
HPV Vaccination up to Date
Participants up to date on HPV vaccination at 1 year post index date.
Time frame: 1 year post index date
Population: Young adults ages 18-26 with a visit at a study clinic during the accrual period, and not up to date on HPV vaccination.
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| CP-CDS | HPV Vaccination up to Date | 90 Participants |
| CP-CDS + SDMT | HPV Vaccination up to Date | 69 Participants |
| Usual Care | HPV Vaccination up to Date | 126 Participants |
Comparison: Generalized linear mixed models with a logit link and binomial error distribution were used to test the effect of the intervention on completion of the HPV vaccination series by 12 months post-index date. The model included two primary contrasts: CP-CDS vs. UC and CP-CDS+SDMT vs. UC and included pre-specified covariates and variables used in the randomization process. Tests of the two planned contrasts were 2-sided with P-values less than 0.025 considered statistically significant.p-value: 0.025Mixed Models Analysis
Comparison: Generalized linear mixed models with a logit link and binomial error distribution were used to test the effect of the intervention on completion of the HPV vaccination series by 12 months post-index date. The model included two primary contrasts: CP-CDS vs. UC and CP-CDS+SDMT vs. UC and included pre-specified covariates and variables used in the randomization process. Tests of the two planned contrasts were 2-sided with P-values less than 0.025 considered statistically significant.p-value: 0.025Mixed Models Analysis
Secondary
Lung Cancer Screening up to Date
Participants up to date on lung cancer screening at 1 year post index date.
Time frame: 1 year post index date
Population: Adult patients with a visit at a study clinic and not up to date for lung cancer screening at index date.
| Arm | Measure | Value (COUNT_OF_PARTICIPANTS) |
|---|---|---|
| CP-CDS | Lung Cancer Screening up to Date | 200 Participants |
| CP-CDS + SDMT | Lung Cancer Screening up to Date | 104 Participants |
| Usual Care | Lung Cancer Screening up to Date | 198 Participants |
Comparison: Generalized linear mixed models with a logit link and binomial error distribution were used to test the effect of the intervention on completion of lung cancer screening by 12 months post-index date. The model included two primary contrasts: CP-CDS vs. UC and CP-CDS+SDMT vs. UC and included pre-specified covariates and variables used in the randomization process. Tests of the two planned contrasts were 2-sided with P-values less than 0.025 considered statistically significant.p-value: 0.025Mixed Models Analysis
Comparison: Generalized linear mixed models with a logit link and binomial error distribution were used to test the intervention effect on the completion of lung cancer screening by 12 months post-index. The model included two primary contrasts: CP-CDS vs. UC and CP-CDS+SDMT vs. UC and included pre-specified covariates and variables used in the randomization process. Tests of the two planned contrasts were 2-sided with P-values less than 0.025 considered statistically significant.p-value: 0.025Mixed Models Analysis