Abstract
Background: Many women fail to adhere to Papanicolaou smear screening guidelines. Although many interventions have been developed to increase screening, the effectiveness of different types of interventions is unclear.
Methods: We performed a systematic review of interventions to increase Papanicolaou smear use published between 1980 and April 2001 and included concurrently or randomized controlled studies with defined outcomes. Interventions were classified as targeted to patients, providers, patients and providers, or health care systems and as behavioral, cognitive, sociologic, or a combination based on the expected action of the intervention. Effect sizes and 95% confidence intervals were calculated for each intervention.
Results: Forty-six studies with 63 separate interventions were included. Most interventions increased Papanicolaou smear use, although in many cases the increase was not statistically significant. Behavioral interventions targeted to patients (eg, mailed or telephone reminders) increased Papanicolaou smear use by up to 18.8%; cognitive and sociologic interventions were only marginally effective, although a single culturally specific, sociologic intervention using a lay health worker increased use by 18.0% (95% confidence interval [CI]: 7.6, 28.4). Provider-targeted interventions were heterogeneous. Interventions that targeted both patients and providers did not appear to be any more effective than interventions targeted to either patients or providers alone. One of the most effective interventions, which introduced a system change by integrating a nurse-practitioner and offered same-day screening, increased screening by 32.7% (95% CI: 20.5, 44.9).
Conclusions: Overall, most interventions increased Papanicolaou smear use, although there was tremendous variability in their effectiveness. Selection of intervention strategies will depend on provider and patient population characteristics and feasibility of implementation.
Screening with regular Papanicolaou smears can decrease not only cervical cancer mortality but also the incidence of invasive disease.1 ,2 Despite increases in recent Papanicolaou smear use during the past two decades,3 ,4 with national estimates of approximately 80% within the past 3 years,4 some women still fail to adhere to recommended Papanicolaou smear screening guidelines,3–26 are found to have advanced disease, and die of invasive cervical cancer.27–29 Thus, the potential benefits of routine screening are not being fully realized. Increasing the provision of Papanicolaou smear counseling by primary care providers and, ultimately, routine Papanicolaou smear use are important components of current Healthy People 2010 goals for reducing cervical cancer mortality.30
Physician recommendation is one of the strongest predictors of screening use,13 ,17 ,31–37 but in many cases, women report that their provider did not recommend Papanicolaou smears.5 ,7 ,18 ,38–45 Explanations for this behavior include lack of time, busy schedules and forgetfulness,46 ,47 beliefs about screening efficacy in the absence of symptoms or concern about proficiency,31 ,48 or confusion about conflicting professional recommendations.46 In the absence of a physician recommendation, patients might assume that screening is unnecessary.
Women who do not receive regular Papanicolaou smears are more likely to be older,5–26 ,49–52 uninsured or underinsured,* lack a usual source of care or regular provider,† and have lower educational attainment or lower household income.‡ Among immigrant women, those who do not speak English or have lower levels of acculturation are also less likely to receive regular Papanicolaou smears.16 ,22 ,23 ,52 ,58 ,64 ,65
Additionally, women might not know about the potential benefits of screening, believe it to be unnecessary in the absence of symptoms,§ fear a potential cancer diagnosis, or believe that cancer cannot be cured.‖ Women might also be concerned about inconvenience, discomfort, embarrassment, or pain associated with the test itself.7 ,25 ,42–45 ,52 ,69 Concerns about modesty or cultural restrictions on gynecological examinations by male physicians could exacerbate other barriers to screening.22 ,25 ,43 ,66 ,72–74 Finally, women might know about the need for Papanicolaou smears but have put off receiving a test because of convenience, time constraints, or forgetfulness.5 ,25 ,38 ,40 ,44 ,59 ,69
During the past two decades, results from numerous interventions to increase Papanicolaou smear use have been reported.75 ,76 These interventions focus on increasing rates of provider recommendation, reducing patient barriers to screening, or both. Two reviews provided descriptive information about interventions published before 1998 and whether they were effective75 ,76 but did not provide details about the magnitude of effectiveness. Because of the large numbers of interventions, particularly in the past several years, and differences in study design, populations, and intervention content, it is difficult to know which interventions are most effective in increasing Papanicolaou smear use. This systematic review of the published literature was conducted to update previous reviews74 ,75 and provide information on the effectiveness of controlled interventions to increase Papanicolaou smear use.
Methods
Study Selection
We used the Ovid search mechanism for MEDLINE to select English language articles published between 1980 and April 2001 on interventions to increase Papanicolaou smear use. The start date of 1980 was chosen because the National Institutes of Health Consensus Conference endorsed routine Papanicolaou smear screening and began to publicize their screening guidelines in that year.77 The search strategy used the subject terms “health behavior,” “patient compliance,” “patient acceptance of health care,” “attitude to health,” “health education,” or “health promotion” (N = 75,985) with the subject terms “Papanicolaou smear” or “cervical neoplasms/prevention and control” (N = 1,871). The combination yielded 467 studies. Study abstracts were reviewed for evidence of random or concurrent assignment of subjects to an intervention or control group, prospective follow-up, and Papanicolaou smear use or recommendation as an outcome. Pre-post designs without control groups were excluded because secular trends to increased Papanicolaou smear use in the past two decades3 ,4 would limit the interpretation of intervention effectiveness. Non-US studies were excluded because differences across systems of care might limit the generalizability of outcomes.
From the review of abstracts, 66 studies were potentially eligible for inclusion. Of this number, 14 were eliminated because they lacked concurrent or randomly assigned control groups and 27 were eliminated because the outcome was not receipt or recommendation of Papanicolaou smear and were non-US studies. Physician-targeted interventions with a documented outcome of recommendation of Papanicolaou smear use were not excluded because these interventions were designed to improve this behavior. Nine interventions were excluded because they were designed to improve follow-up after abnormal Papanicolaou smears, which left 16 studies as a result of the search.78–90 Because electronic searches might not identify all relevant studies,91 reference lists and published reviews of interventions70 ,75 ,76 were reviewed, and a hand search of the journals Preventive Medicine and American Journal of Preventive Medicine for 1999 through April 2001 was conducted to find other eligible studies. Thirty additional studies were found,92–124 and 46 studies are included in this article.
Data Abstraction
Data were abstracted from studies using a standardized abstraction format to describe the intervention target (ie, patient, physician, patient and physician, or health care system), type of intervention, intervention content, means of determining receipt of Papanicolaou smear (eg, self-report, chart), characteristics of the patient population, and intervention effectiveness. In randomized trials where the sample size was not reported for each arm, the sample was divided by the number of study arms to estimate the number of women in each arm.
Each intervention within a study was abstracted separately. Within studies, interventions were classified based on the underlying mechanism of an intervention to increase Papanicolaou smear use using a previously developed classification scheme.125 Interventions were classified as behavioral, cognitive, sociologic, or a combination of the three. Behavioral interventions change stimuli associated with Papanicolaou smear use (eg, reminders). Cognitive interventions provide new information, educate women about Papanicolaou smears, and clarify any existing misperceptions. Cognitive interventions were categorized further as individually tailored or based on theory (eg, health belief model) and as using generic educational materials. Cognitive interventions were also classified based on the method of delivery—interactively (eg, by telephone or in person) or statically (eg, by letter or pamphlet). Sociologic interventions use social norms or peers to increase Papanicolaou smear use (eg, lay health workers, peer counselors).
Interventions were classified further based on the content of the intervention (eg, telephone reminder) and type of control group to which the intervention was compared. Studies that included a lower level intervention to increase Papanicolaou smear use as the control group (eg, postcard reminder) were considered to have active controls. Studies in which the control group did not receive any specific strategies to increase Papanicolaou smear use were classified as having usual-care controls. In studies with combined interventions in which the same strategy was applied to intervention and control groups (active controls), the intervention was categorized by the difference between intervention and control groups.
Data Analysis
Effect sizes and 95% confidence intervals (CI) were calculated for each intervention. For randomized studies, intervention effectiveness was calculated as the difference in Papanicolaou smear use between the intervention and control groups at the first assessment of intervention effectiveness. Variance was calculated for binomial proportions for intervention and control groups. For concurrently controlled studies, the effect size was calculated as the difference between the screening rates before and after the intervention for the intervention group and the control group ([Pscreened postintervention–Pscreened preintervention] − [Pscreened postcontrol–Pscreened precontrol]). Variance was calculated for the binomial proportions for intervention and control groups both before and after the intervention.
Compliance scores, which measured compliance with recommended screening frequency (number of screenings during a specified period divided by the number of women eligible for screening), were treated as the proportion of women screened for the purposes of this article. In two studies, compliance scores were based on a 3-year frequency of Papanicolaou smear use108 ,109 and were greater than 100% because most screening guidelines recommend screening more often than every 3 years. These scores were divided by 3 to create an average annual compliance score.
Within each group of intervention (eg, reminders mailed to patients compared with active controls), the effect size and 95% confidence interval for each intervention was graphed, and graphs were inspected visually for signs of heterogeneity. Because of the variability in patient populations and setting, as well as small numbers of interventions in any single category, we did not perform quantitative analysis.
Results
Of the 46 studies, there were 63 separate interventions; 24 were targeted to patients, 25 to physicians, and 12 to both patients and physicians. Two interventions introduced system changes (Table 1 78–90 ,92–124). Most studies were randomized controlled trials (n = 31; 67.4%). Overall, the largest number of interventions, 22, used behavioral strategies; all other categories had 10 or fewer interventions each. Most studies included women in the 50- to 59-year age-group (60.9%), and approximately 30% of the studies included women younger than 40 years. Many studies (41.3%) included a sizable portion (>20%) of minority women. About one half of the studies reported health insurance status of the patient population. Among this group, most were conducted in samples where at least 75% of the patient population had health insurance. Finally, the period of assessment following intervention delivery was most frequently 12 or more months, although many studies also reported assessments of 3 to 6 and 6 to 12 months.
The types of interventions changed during the past two decades. In general, most interventions conducted in the 1980s and early 1990s were compared with usual-care controls, whereas those conducted in the 1990s were mainly compared with active controls. Additionally, interventions conducted in the late 1990s generally included multiple different components and targeted patients and providers, whereas those conducted in earlier years focused on single components and targeted patients or providers. For example, in an intervention published in 1999, Paskett and colleagues79 reported the combined effectiveness of an intervention that targeted both patients and providers and used letter reminders, office reminders, multiple educational strategies, mass media, and lay health workers. Earlier interventions evaluated provider reminders alone.119
Patient-Targeted Interventions
The 24 patient-targeted interventions are listed in Table 2.¶
Behavioral Interventions
Compared with usual care, the single behavioral intervention was very effective and increased Papanicolaou smear use by 24.4% (95% CI: 11.1, 37.7).96 Among the five interventions that were compared with active controls, one intervention was associated with decreased rates of Papanicolaou smear use (−8.6%; 95% CI: −13.1, −4.1).111 Compared with active controls, the other four behavioral strategies were effective, and improvements in Papanicolaou smear use ranged from 10.1% to 18.8%.86 ,103 ,116 The telephone reminder was associated with the largest increase in Papanicolaou smear use and was from a single study in which multiple interventions were tested separately.86
Cognitive Interventions
All four cognitive interventions were theory based and compared with active controls. The three delivered statically by letter did not improve Papanicolaou smear use, but the fourth, delivered interactively by telephone, led to an increase in Papanicolaou smear use of 8%, although this finding was not statistically significant (95% CI: −1.4, 17.4).113
Two of the three behavioral and cognitive interventions utilized mailed generic educational information and reminders and were not associated with significant improvements in Papanicolaou smear use.85 ,94 The third, which incorporated a telephone call from a health educator with the reminder in the intervention group, reported a 13.5% increase in Papanicolaou smear use (95% CI: 7.6,19.4).
Sociologic Interventions
Most of the sociologic interventions were targeted to specific ethnic groups and used lay health workers,126 women recruited from the target population, to encourage women to attend screening.80 ,82–84 ,87 The interventions that used sociologic strategies alone and sociologic and cognitive strategies had similar effects, and most improved Papanicolaou smear use from 2.7% to 9.2%,78 ,82–84 ,88 ,89 ,100 ,122 although not all increases were statistically significant. Two interventions were associated with decreased Papanicolaou smear use,80 ,87 although this finding was not statistically significant. The authors of one these interventions speculated that increased outreach efforts in the comparison community were responsible for a lack of effect.80 Two of the three sociologic, behavioral, and cognitive interventions led to large increases in Papanicolaou smear use, 18.0% and 36.0%.82 ,122 The first, conducted by Bird and colleagues,82 used lay health workers, educational pamphlets, and financial incentives and was culturally specific to Vietnamese-American women. The apparent effectiveness of the second, conducted by Whitman and colleagues,122 might be due to the decline in Papanicolaou smear use from 39% to 17% in the concurrent control group. The other intervention in this category relied on media to present role model behaviors and did not lead to increased Papanicolaou smear use.81
Provider-Targeted and Provider- and Patient-Targeted Interventions
The 27 provider targeted interventions are listed in Table 3,** and the 12 provider- and patient-targeted interventions are listed in Table 4.††
Behavioral Interventions
The nine interventions compared with usual-care controls‡‡ did not appear to have higher rates of screening than similar interventions compared with active controls§§ (Table 3). There was great heterogeneity among the behavioral interventions, ranging from a statistically significant 18% decrease in Papanicolaou smear use93 to a 44% increase.90 There was no clear difference in effectiveness between chart reminders and flow sheets requiring completion; in fact, both studies with extreme findings used flow sheets requiring completion. In the study with decreased Papanicolaou smear use, the authors suggested that the intervention was incompletely implemented, leading to an unexpected decline in Papanicolaou smear use. The same intervention, however, was also assessed for other preventive strategies and was associated with increased clinical breast examination, mammography, and three types of immunizations.93 The intervention associated with a 44% increase in Papanicolaou smear use was based on a potentially inappropriate concurrent control group—preintervention rates of Papanicolaou smear use were 98% in the control group compared with 43% in intervention arm.92
Provider- and patient-targeted behavioral interventions were also heterogeneous and ranged in effectiveness from a 6% decrease (95% CI:−10.5, −1.5)111 to 13.8% increase (95% CI:10.9, 16.7) in Papanicolaou smear use116 (Table 4).
Cognitive Interventions
The three interventions that used seminars or audit with feedback97 ,108 ,119 led to slight increases in Papanicolaou smear use, from 2%97 to 8%.108 Three interventions used a combination of cognitive and behavioral strategies.97 ,114 ,119 The effects were varied and ranged from a significant decline in Papanicolaou smear use of −6% (95% CI: −11.9, −0.1)119 to an 18% increase (95% CI: −21.3, 57.3)114 (Table 3).
Interventions targeted to patients and physicians using behavioral and cognitive strategies were also variable in their impact on Papanicolaou smear use, ranging from a 7% decrease in use (95% CI: −10.6, −3.4)98 to a 13% increase in Papanicolaou smear use (95% CI: −7.5, 32.5).121 Although most estimates were positive, only one, which included six separate intervention components, was statistically significant106 (Table 4).
Sociologic Interventions
The combined sociologic, behavioral, and cognitive intervention strategy targeted to physicians and patients that used a generic educational strategy had little impact on Papanicolaou smear use,123 whereas the intervention that used church liaisons, educational mass media campaigns, lay health workers, theory-based education, and community activities as the cognitive strategies79 ,127 led to a 21% increase in Papanicolaou smear use (95% CI: 6.0, 36.0) (Table 4).
System Interventions
Finally, two interventions introduced system changes.105 ,117 One altered how medical care was delivered by integrating into a clinic a nurse-practitioner who was able to perform same-day screening.105 This intervention was extremely effective, leading to a 32.7% increase in Papanicolaou smear use (95% CI: 20.5, 44.9). The other system intervention used trained community health workers and led to a 7% increase in Papanicolaou smear use (95% CI: −1.3, 15.3), although it was not statistically significant.
Discussion
As shown in this systematic review of interventions to increase Papanicolaou smear use, many patient and provider barriers can be overcome with well-implemented interventions. Selection and implementation of intervention strategies to improve Papanicolaou smear screening can be based, in part, on the characteristics of the underlying patient and provider populations. For example, patient forgetfulness can be overcome with behavioral reminders sent to the home or delivered by telephone. Telephone reminders increased Papanicolaou smear use by up to 18% (95% CI: 15.8, 21.8).86 Such reminder interventions might be particularly effective in increasing regular screening among women who previously had a Papanicolaou smear. Forgetfulness in providers can also be overcome with behavioral reminders included on the chart or flow sheets requiring completion, although these findings varied widely.
Women who have not had a previous Papanicolaou smear might be difficult to reach through traditional medical settings and might require more extensive outreach strategies. Sociologic strategies, which use lay health workers, could be effective in populations who might distrust the health care system or have other cultural barriers to screening. Sociologic strategies were marginally effective in increasing Papanicolaou smear use, although a culturally sensitive intervention that also included behavioral and cognitive strategies was particularly effective in a population of Vietnamese-American women, increasing Papanicolaou smear use by 18.0% (95% CI: 7.6, 28.4).82
Few interventions focused on patient lack of knowledge or fears of Papanicolaou smear, although use of interactive delivery of cognitive educational interventions by telephone was associated with increased Papanicolaou smear use. Provider educational strategies developed to clarify information about screening efficacy and guidelines or to improve proficiency were marginally effective in increasing Papanicolaou smear use. Interventions that targeted both patients and providers appeared to increase screening, although they were generally not much more effective than similar interventions targeted to either patients or providers alone.
Few interventions introduced system level changes, although one that integrated a nurse practitioner into a primary care practice who could also perform same-day screening led to a large increase in Papanicolaou smear use (32.7%; 95% CI: 20.5, 44.9).105 Other uncontrolled strategies that use same-day screening in emergency departments128 ,129 or churches130 have been reported to be feasible and to increase Papanicolaou smear use. These opportunistic screening strategies might also be more resource intensive and require reservation of personnel and facilities for performance of screening on demand. Other effective strategies, such as patient or provider reminders, could be inexpensive on an on-going basis, but might require an initial investment in computer infrastructure. Thus, measuring the cost and cost-effectiveness of different intervention strategies will provide important information for health departments, providers, and other health care organizations about the feasibility of implementing interventions to improve Papanicolaou use.38 ,131
In several cases, interventions were ineffective in increasing Papanicolaou smear use, but the same intervention in the same provider and patient populations led to large increases in mammography or other preventive services.93 ,119 There are several potential alternative explanations for this finding, including differences in a woman’s perception of these tests or likelihood of developing cancer, differences in provider beliefs about proficiency, or the time required for performance of different screening tests. Exploring potential differences in intervention effectiveness and any differential impact of the underlying barriers for screening in different types of screening tests will be an important area for future cancer control research.
There are some limitations with this review, including the reliance on an electronic search of the published literature for identifying interventions, the combination of multiple measures of Papanicolaou smear utilization (eg, self-report, chart review) or documented Papanicolaou smear ordering, and discrepancies between the unit of randomization and the unit of analysis in published interventions. Our search strategy, although similar to one used successfully for meta-analyses of interventions to increase mammography use,132 ,133 found fewer interventions than did the review of bibliographies or the hand search of recent articles published in prevention-oriented journals. Additionally, studies with negative findings might be less likely to be published91 and found by our search strategy. As a result, the estimates of intervention effectiveness presented here might overstate their true effectiveness.
Studies included here used different mechanisms to determine Papanicolaou smear use, including self-report, chart-audit, and electronic claims. Papanicolaou smear self-report has been reported to overestimate utilization when compared with charts or claims data, with reports of accuracy ranging from 67% to 99% agreement.134–138 Similarly, the timing of the measurement of Papanicolaou smear receipt differed across the studies, with some studies reporting Papanicolaou smear use within 6 months of the intervention‖‖ and others reporting Papanicolaou smear use 1 year or more 2 years or more after the intervention was initiated.¶¶ Additionally, some of the provider-targeted interventions were based on whether Papanicolaou smears were recommended or ordered,95 ,99 ,101 ,107 ,118 rather than on a woman’s receipt of Papanicolaou smear. Within a given study, however, women in the intervention and control arms should be equally likely to overstate utilization or not comply with provider recommended Papanicolaou smear, so the relative estimate (intervention-control) is unlikely to be affected. In two studies, Papanicolaou smear use was measured as a 3-year compliance rate and converted to an annual rate because compliance was greater than 100%.108 ,109 These annual rates might understate the effectiveness of these interventions.
Several studies randomized clinics, physician practices, churches, or more broadly defined communities to intervention and control conditions and then performed analysis on the number of women in each of the groups, rather than the unit of randomization.92 ,111 Women living within a specific community or treated at a similar clinic are more likely to have similar behaviors and are not independent observations.139 If the actual unit of randomization or the correlation among women was accounted for in analysis, the point estimate of intervention effectiveness would not be affected, but the confidence interval would likely be wider. As a result, the confidence intervals reported here could overestimate the effectiveness of interventions to increase Papanicolaou smear use.
Notes
This work was supported in part by the National Cancer Institute contract (KRY and JM) and NIA grant RO1-AG15340 (JM).
An earlier version of this work was prepared for the National Cancer Institute report entitled “Disparities in Cervical Cancer Outcomes: What Do We Know? What Do We Need to Know?”
The views in this paper reflect those of the authors and not the National Cancer Institute.
↵* References 5 ,6 ,13 ,16 ,17 ,19 ,23 ,24 ,26 ,42 ,43 ,45 ,51 ,53–55.
↵† References 17 ,18 ,22 ,32 ,34 ,36 ,40 ,41 ,45 ,54 ,56 ,57.
↵‡ References 5 ,6 ,8 ,14–16 ,18–20 ,22 ,23 ,40 ,41 ,49–51 ,54 ,58–63.
↵¶ References 78 ,80–89 ,94 ,96 ,100 ,102 ,103 ,111–113 ,116 ,122.
↵** References 85 ,86 ,90 ,92 ,93 ,95 ,97 ,101 ,104 ,107 ,108 ,111 ,114 ,115 ,118–121 ,124.
↵†† References 79 ,85 ,98 ,99 ,106 ,109–111 ,116 ,121 ,123 ,124.
- Received for publication July 16, 2002.
- Revision received July 16, 2002.