Navigators can facilitate timely access to cancer services, but to the authors' knowledge there are little data available regarding their economic impact.
Navigators can facilitate timely access to cancer services, but to the authors' knowledge there are little data available regarding their economic impact.
The authors conducted a cost-consequence analysis of navigation versus usual care among 10,521 individuals with abnormal breast, cervical, colorectal, or prostate cancer screening results who enrolled in the Patient Navigation Research Program study from January 1, 2006 to March 31, 2010. Navigation costs included diagnostic evaluation, patient and staff time, materials, and overhead. Consequences or outcomes were time to diagnostic resolution and probability of resolution. Differences in costs and outcomes were evaluated using multilevel, mixed-effects regression modeling adjusting for age, race/ethnicity, language, marital status, insurance status, cancer, and site clustering.
The majority of individuals were members of a minority (70.7%) and uninsured or publically insured (72.7%). Diagnostic resolution was higher for navigation versus usual care at 180 days (56.2% vs 53.8%; P = .008) and 270 days (70.0% vs 68.2%; P < .001). Although there were no differences in the average number of days to resolution between the 2 groups (110 days vs 109 days; P = .63), the probability of ever having diagnostic resolution was higher for the navigation group versus the usual-care group (84.5% vs 79.6%; P < .001). The added cost of navigation versus usual care was $275 per patient (95% confidence interval, $260-$290; P < .001). There was no significant difference in stage distribution among the 12.4% of patients in the navigation group vs 11% of the usual-care patients diagnosed with cancer.
Navigation adds costs and modestly increases the probability of diagnostic resolution among patients with abnormal screening test results. Navigation is only likely to be cost-effective if improved resolution translates into an earlier cancer stage at the time of diagnosis. Cancer 2014;120:570–578. © 2013 American Cancer Society.
Despite improvements in cancer care, there are persistent disparities in stage of disease at the time of presentation, access to treatment, and survival between minority and socioeconomically disadvantaged populations compared with their white and/or more advantaged counterparts.[1, 2] Inadequate follow-up after an abnormal screening test and/or a cancer diagnosis are potential contributors to these disparate outcomes because < 75% of disadvantaged individuals with an abnormal cancer screening examination receive timely diagnostic care.[3-6] Patient navigation programs were developed to reduce such gaps in care by improving access to, and timeliness of, cancer services.[7, 8]
Although patient navigation programs hold promise as a means of reducing disparities in cancer screening follow-up, to the best of our knowledge, the evidence supporting their efficacy in decreasing mortality or economic impact is limited. Studies demonstrating that navigation programs increase participation in cancer screening and adherence to diagnostic follow-up have lacked control groups or have had relatively small sample sizes.[8, 9] To our knowledge, only a single modeling-based economic evaluation of navigation for breast cancer has been published to date.
The Patient Navigation Research Program (PNRP) study was designed to examine the benefits of navigation for minority/underserved patients with abnormal breast, cervical, colorectal, or prostate cancer screening results.[11, 12] We conducted an economic evaluation alongside the PNRP to estimate the short-term costs and outcomes of navigation from receipt of an abnormal screening examination to diagnostic resolution (benign or cancer).
The PNRP was a multisite prospective study that evaluated the impact of navigation for individuals aged 18 years to 98 years from underserved populations with abnormal screening results for breast, prostate, colorectal, or cervical cancer. Underserved populations included African Americans, Hispanics, Asians, and American Indians/Alaska Natives or low-income populations. The study was conducted at 9 sites between January 1, 2006 and March 31, 2010. Although all sites shared a common definition of patient navigation, common methods for data collection, enrollment, and follow-up periods, each implemented navigation within the real-world context of their community and health care setting. This resulted in varied research designs for comparing navigation with usual care: 2 sites used an individually randomized trial design, 2 sites used a group-randomized trial design, and 5 sites used quasiexperimental designs with nonrandom allocation into groups. This approach allowed for the needs of community-based participatory research, as well as the traditional multicenter clinical trials analysis.[13-19] Usual care consisted of standard diagnostic follow-up for an abnormal screening result without navigator support. Navigation added the provision of support and guidance for timely access to the cancer care system, addressing barriers, and facilitating quality care.[11, 20] Navigators were professional health workers and/or lay persons (eg, cancer survivors or community volunteers). In addition to site-specific training, all navigators participated in a national training program .
We conducted an economic evaluation of navigation versus usual care using the societal perspective, including health sector and patient costs.[23-25] We focused on the primary PNRP study outcome: time to diagnostic resolution. Because the time horizon of the PNRP was < 365 days for 92% of participants, we only consider events occurring within the study period without discounting future costs or effects. Results are presented in a cost-consequence format[12, 27] in 2011 US dollars (USD). All data collection procedures were approved by the respective Institutional Review Boards of the PNRP sites.
Data regarding direct medical resources for diagnostic follow-up tests and services after an abnormal screening result were based on PNRP study records. To estimate the fixed and variable costs related to providing patient navigation services,[12, 29] we surveyed PNRP program managers regarding the resources used to establish and maintain each navigation program. Information from navigator tracking logs was used to measure the average time spent providing navigation from the time of an abnormal screening result to diagnostic resolution (cancer or noncancer) in the periods during which the program was in a steady state (ie, excluding study start-up and training periods).
Tests and services were matched to their corresponding Healthcare Common Procedure Coding System codes and valued based on 2011 Medicare fee schedules published by the Centers for Medicare and Medicaid Services.[31, 32]
To estimate the cost of establishing navigation services, the cost of navigator recruitment was based on time for secretarial staff to produce and post advertisements and for managers to review resumes (the cost of the advertisements were excluded). Navigator training costs included time for adapting national PNRP navigator training and teaching materials to the local site and the time for navigators and their direct supervisor(s) to attend training. Based on the PNRP training standards, we assumed supervisors received 1 day of training and navigators received 3 days of training. Although PNRP training was provided at no cost by the American Cancer Society in partnership with the National Cancer Institute, we estimated that training costs were $100 per day. Initial recruitment and training costs were annuitized over the working half-life of navigators. The cost to purchase office furnishings and equipment was converted to 2011 USD and then applied as a 5-year annuitized cost for the duration of the study period at a discount rate of 10%.
The estimated fixed cost to maintain navigation services included the cost of additional office space for navigation staff valued using the published national average adjusted to 2011 USD. Costs associated with navigator travel (eg, taxi, bus, and train fares) and navigation materials (eg, office supplies and telephone and mail charges) were based on costs reported by program managers. Navigator mileage was valued based on the Internal Revenue Service standard mileage rate for 2011. Time required for supervising, providing administrative support to navigators, and receiving ongoing training were estimated based on average time requirements reported by program managers and valued using the respective national wage rate. The direct cost of providing navigation services was based on staff time recorded in navigator logs based on median times for the 15-minute time intervals recorded. Times to visit termination were recorded per patient and were truncated at 90 minutes for scheduling time and at 240 minutes for direct care for all patients. Navigation services provided by professional navigators were valued using median national wage rates for registered nurses. Lay navigator rates were based on the median hourly wage for nursing aides, orderlies, and attendants. Services that used a mix of professional and lay navigators used the average of the 2 wage groups. Wage rates were inflated to 2011 USD using the consumer price index medical services component with a 30% fringe rate.
Navigator turnover observed in the PNRP study was 26% per year. We valued the cost of replacing navigators based on estimated recruitment and training costs plus the cost of lost productivity.
Patient time associated with seeking care included time spent traveling to health care services and time spent using these services. Information regarding patient home and clinic zip codes and corresponding centroid data were used to calculate the approximate distance traveled to receive services. Travel costs were estimated using the Internal Revenue Service standard mileage rate for 2011. Time spent obtaining medical care related to the abnormal screening test was estimated based on expert clinical opinion plus a 20% wait time. Time was valued using PNRP census region-specific wage rates from the Bureau of Labor Statistics for individuals matched by their age and sex.
The primary PNRP study outcome was time from abnormal screening test result to diagnostic resolution (cancer or noncancer). This was analyzed via a meta-analysis comparing adjusted proportional hazards ratios of patient navigation versus usual care for each cancer type across each site. A 2-part approach was used to estimate: 1) the mean number of days to resolution among those who were observed through to the point of diagnostic resolution, and 2) the probability of patients completing follow-up by different time points (60 days, 90 days, 180 days, 270 days, and ever). We also performed subgroup analyses evaluating the mean number of days to diagnostic resolution for those patients diagnosed with cancer and those diagnosed with early-stage neoplasia (cervical intraepithelial neoplasia type 2 [CIN2]/CIN3/stage 0 and stage I) versus more advanced cancers (TNM stage II to stage IV).
Categorical patient characteristics were compared using independent chi-square tests. We tested for differences in total costs and consequences between navigation versus usual care[23, 24, 26] using multilevel mixed-effects linear regression modeling. Consequences included the mean probability of resolution by 60 days, 90 days, 180 days, 270 days, and ever (yes vs no) and being diagnosed with early-stage versus late-stage disease among those with cancer. Because the navigated and usual-care groups were not balanced, all analyses were adjusted for an a priori set of defined covariates including age (in decades), race/ethnicity (African American, white, Hispanic, or other), primary language spoken (English, Spanish, or other), marital status (never married, married, or other), insurance status (uninsured, public insurance, or private insurance), and cancer site. Clustering was accounted for at the site level by including site as a random effect in the mixed-effects models. To incorporate uncertainty surrounding the estimates of the cost of providing navigation services, we used results from probabilistic modeling with 1000 simulations for each site in all analyses.
Less than 5% of item-level patient survey information was missing; patients were missing <10% of data for demographic or clinical characteristics. Missing survey data were replaced using the Schonlau implementation of the hot deck multiple imputation procedure and missing patient data were imputed using chained equations.
The cost analysis and probabilistic modeling were completed in Microsoft Excel 2007 (Microsoft Corporation, Redmond,Wash). Zip code centroid data and distance approximation used SAS statistical software (version 9.2; SAS Institute Inc, Cary, NC) and statistical analyses were completed in Stata statistical software (version 12.1; StataCorp, College Station, Tex).
The PNRP included 10,521 patients (5063 patients in the navigation group and 5458 in the usual-care group) (Table 1). A total of 1253 patients (12%) were excluded from the days to resolution analysis because they did not achieve resolution before the end of study follow-up (365 days) or were missing time to resolution data (<1%) (Fig. 1). A high percentage of participants were either Hispanic (39%) or African American (32%). Approximately 40% were publically insured. Only 11% of patients in the usual-care group and 12.4% of navigated patients were diagnosed with cancer.
|Characteristics||Total % (No.)||Navigation % (No.)||Usual Care % (No.)||P|
|100 (10,521)||48.1 (5063)||51.9 (5458)|
|<30||18.6 (1958)||22.0 (1115)||15.4 (843)||<.001|
|30-39||13.8 (1451)||16.6 (841)||11.2 (610)|
|40-49||28.4 (2988)||26.5 (1340)||30.2 (1648)|
|50-59||20.4 (2146)||19 (950)||21.9 (1196)|
|60-69||12.5 (1313)||11 (560)||13.8 (753)|
|70-79||4.8 (507)||4 (217)||5.3 (290)|
|≥80||1.2 (125)||0.5 (27)||1.8 (98)|
|Missing data||0.3 (33)||0.3 (13)||0.4 (20)|
|Female||91.9 (9671)||92.1 (4665)||91.7 (5006)||.471|
|Male||8.1 (849)||7.9 (398)||8.3 (451)|
|Missing||<0.1 (1)||0 (0)||<0.1 (1)|
|African American||31.7 (3330)||29.4 (1,487)||33.8 (1843)||<.001|
|White||24.7 (2594)||24.2 (1224)||25.1 (1370)|
|Hispanic||39 (4106)||42.3 (2142)||36 (1964)|
|Other||3.7 (392)||4.1 (207)||3.4 (185)|
|Missing data||0.9 (99)||0.1 (3)||1.8 (96)|
|English||59.7 (6286)||62.4 (3159)||57.3 (3127)||<.001|
|Spanish||24.5 (2576)||29.9 (1515)||19.4 (1061)|
|Other||5.6 (594)||6.2 (312)||5.2 (282)|
|Missing data||10.1 (1065)||1.5 (77)||18.1 (988)|
|Single/never married||41 (4317)||43.1 (2180)||39.2 (2137)||<.001|
|Married/living as married||31.9 (3360)||35 (1772)||29.1 (1588)|
|Divorced/separated||13.3 (1397)||15.5 (784)||11.2 (613)|
|Widowed||4.3 (457)||3.9 (197)||4.8 (260)|
|Missing data||9.4 (990)||2.6 (130)||15.8 (860)|
|Uninsured||32.2 (3385)||36.3 (1837)||28.4 (1548)||<.001|
|Public||40.5 (4259)||38.9 (1969)||42 (2290)|
|Private||26.6 (2801)||23.7 (1202)||29.3 (1599)|
|Missing data||0.7 (76)||1.1 (55)||0.4 (21)|
|Cancer screening test|
|Breast||63.9 (6726)||60.9 (3083)||66.7 (3643)||<.001|
|Cervix||25.5 (2681)||28.7 (1455)||22.5 (1226)|
|Colorectal||4.7 (497)||4.3 (219)||5.1 (278)|
|Prostate||5.9 (617)||6 (306)||5.7 (311)|
|No cancer||76.5 (8050)||77.7 (3934)||75.4 (4116)||<.001|
|Cancer||11.7 (1226)||12.4 (626)||11 (600)|
|Missing data||11.8 (1245)||9.9 (503)||13.6 (742)|
|Stage of disease (TNM)|
|0||11.5 (141)||7.5 (47)||15.7 (94)||<.001|
|I||18.9 (232)||18.1 (113)||19.8 (119)|
|II||20.7 (254)||19.6 (123)||21.8 (131)|
|III||7.3 (89)||7.8 (49)||6.7 (40)|
|IV||2.9 (35)||3.2 (20)||2.5 (15)|
|CIN2 or CIN3||25.4 (311)||29.6 (185)||21 (126)|
|Missing data||13.4 (164)||14.2 (89)||12.5 (75)|
The average cost to hire and train a patient navigator was $2460; the ongoing maintenance cost for each navigator, exclusive of wages, was $24,140 annually (Table 2). Participants in the navigation group received >14,000 hours of navigation at an estimated average total cost of $190 per patient (Table 3).
|Cost||Mean (SE) 2011 USDa|
|Total cost||2460 (180)|
|Office space||5160 (840)|
|Administrative support||2060 (440)|
|Office supplies||420 (130)|
|Parking and travel||1070 (610)|
|Ongoing training||670 (210)|
|Annualized total cost||24,140 (6350)|
|Screening||No. Navigated to Diagnostic Resolution||Total Navigation Time||Average Navigation Time Per Person||Average Value of Navigator Time Per Person, Mean (95% CI)a||Average Total Costs Per Person, Mean (95% CI)b|
|Test||% (No.)||% (No.), Hours||Hours||2011 USD||2011 USD|
|Breast||57 (2468)||73 (10,344)||4.2||150 (50-460)||210 (100-560)|
|Cervical||31 (1358)||14 (2043)||1.5||70 (40-120)||130 (60-230)|
|Colorectal||5 (214)||9 (1220)||5.7||210 (140-450)||300 (180-530)|
|Prostate||7 (290)||4 (508)||1.8||110 (70-190)||200 (160-260)|
|Total||100 (4330)||100 (14,114)||3.3||130 (60-450)||190 (80-540)|
The average value of navigator time per person, including wages and benefits, was $130 (95% confidence interval [95% CI], $60-$450). Navigated patients had higher adjusted mean diagnostic follow-up costs ($400 vs $320) and patient time costs ($70 vs $18) compared with patients receiving usual care. Considering all costs, the total adjusted incremental cost of navigation versus usual care was $275 (95% CI, $260-$290) (Table 4). Unadjusted results and results excluding imputed cost data were found to have little impact on this result (mean, $284; 95% CI, $265-$300).
|Navigation||Usual Care||Incremental Change||P|
|Costs, 2011 USD (95% CI)|
|Navigation costs||$190 ($80-$540)||—||—||—|
|Medical care costs|
|Diagnostic follow-up||$400 ($380-$410)||$320 ($260-$380)||$80 ($60-$90)||<.001|
|Time||$70 ($66-$72)||$18 ($2-$34)||$51 ($48-$54)||<.001|
|Travel||$7 ($5-$9)||$8 ($3-$12)||$-1 ($-3 to $1)||.191|
|Total cost||$635 ($620-$650)||$360 ($260-$460)||$275 ($260-$290)||<.001|
|Mean no. of d to diagnostic resolution (95% CI)|
|110 (106-115)||109 (90-128)||1.0 (−3 to 6)||.630|
|Probability of diagnostic resolution, mean (95% CI)|
|By 60 d||43.6% (41.7%−45.5%)||42.3% (34.4%−50.2%)||1.3% (−0.5% to 3.2%)||.165|
|By 90 d||52.3% (50.4%−54.1%)||51.0% (42.1%−59.9%)||1.3% (−0.6% to 3.1%)||.195|
|By 180 d||56.2% (53.2%−57.9%)||53.8% (44.2%−63.3%)||2.4% (0.6%−4.1%)||.008|
|By 270 d||70.0% (60.7%−71.5%)||67.2% (57.9%−76.6%)||2.8% (1.3%−4.3%)||<.001|
|Ever vs never||84.5% (83.3%−85.7%)||79.6% (75.5%−83.7%)||4.9% (3.7%−6.2%)||<.001|
|Patients diagnosed with cancer: d to diagnostic resolution, mean (95% CI)|
|106 (96-116)||104 (49-158)||2.0 (−8 to 12)||.666|
|Patients diagnosed with cancer: stage at diagnosis, mean (95% CI)|
|Early vs lateb||57.9% (52.7%−63.1%)||52.2% (35.7%−68.7%)||5.7% (0.5%−10.9%)||.074|
|Patients diagnosed with early stage cancer: d to diagnostic resolution, mean (95% CI)|
|90 (79-102)||86 (55-117)||4.0 (−7 to 16)||.468|
|Patients diagnosed with late-stage cancer: d to diagnostic resolution, mean (95% CI)|
|91 (68-113)||81 (7-153)||10.0 (−13 to 32)||.416|
Navigation increased the probability of diagnostic resolution after 180 days and 270 days. The adjusted probability of ever achieving diagnostic resolution was higher in the navigation group compared with the usual-care group (84.5% vs 79.6%, 4.9% increase; 95% CI, 3.7%-6.2% [P < .001]); unadjusted results were found to be similar (data not shown). Among patients who achieved diagnostic resolution, the adjusted mean time to resolution was 110 days (95% CI, 106 days-115 days) in the navigation group and 109 days (95% CI, 90 days-128 days) in the usual-care group (P = .630) (Table 3). For those patients who were diagnosed with cancer, the adjusted mean time to resolution was 106 days (95% CI, 96 days-116 days) for navigated patients and 104 days (95% CI, 49 days-158 days) for patients in the usual-care group (P = .66). With adjustment, there was also no significant difference found among precancerous and early-stage versus late-stage diagnoses (57.9% in the navigated vs 52.2% in the usual-care group; P = .074), days to diagnosis for patients with early-stage disease (90 days in the navigated vs 86 days in the usual-care group; P = .468), or days to diagnosis for those with late-stage disease (91 days in the navigated vs 81 days in the usual-care group; P = .416). The imputation of missing covariate data was found to have no impact on these results.
To the best of our knowledge, the current study is the first to examine the costs of navigation as well as evaluate its impact in a large national program assessing navigation for largely underserved patients with abnormal cancer screening findings. The results indicate that navigation yields a small but significant increase in the probability of diagnostic resolution after 180 days and 270 days after an abnormal cancer screening test at an added incremental cost of $275 per person compared with usual care. However, the added costs of navigation services did not translate into downstaging of cancer among the 11% to 12% of patients with abnormal test results who were diagnosed with cancer.
The relatively short time frame for the current study prevents an estimation of longer-term outcomes such as cancer mortality. However, the small differences in outcomes and relatively large incremental costs observed for navigated patients suggest that general patient navigation programs among low-income and underserved populations with abnormal cancer screening test results are unlikely to be cost-effective (either expressed as cost per life-year gained or quality-adjusted life-year gained) using commonly cited thresholds for cost-effectiveness (eg, $50,000 or $100,000). In the current study, navigated patients were more likely to be younger, of minority race, not have English as their primary language, single or divorced, and uninsured. If patients with these characteristics were more likely to need and benefit from navigation, this should have biased the study in favor of seeing an effect. When we remove adjustment for these factors, the results were found to differ little from the adjusted results in magnitude or significance (data available upon request).
One previous study investigating the cost-effectiveness of patient navigation extended data from mammography follow-up from a single institution in a decision analysis model. The analysis assumed a 6-month difference in the time to diagnostic resolution between navigation and usual care. Under this optimistic assumption, navigation cost $114,800 per life-year saved (LYS) (adjusted to 2011 USD). If there was a 3-month difference, costs increased to $235,280 per LYS. Navigation was only cost-effective ($43,520 per LYS) under the combination of the most favorable assumptions, including a 6-month earlier time of diagnosis, a 15% higher probability of obtaining follow-up resolution, and the supposition that those lost to follow-up present at more advanced stages of cancer.
Navigation of patients with abnormal cancer screening results may provide value to consumers in other domains. For example, it may lessen the anxiety associated with having an abnormal test result and negotiating the medical care system, enhance patient satisfaction, and/or improve health-related quality of life. Analysis of PNRP patient satisfaction surveys showed no significant difference in the crude satisfaction scores of navigated and usual-care patients.[46, 47] Navigation might also improve the efficiency of clinical services by ensuring that scheduled follow-up appointments are not missed, thereby decreasing gaps in provider productivity. It could also assist patients in obtaining screening coverage through federal or local programs, reducing the burden to health systems of uncompensated care.[48, 49] Positive navigation experiences could also lead to improved adherence to subsequent regular screening.
Given the very modest effects of the navigation program evaluated in the current study, other strategies to improve the timeliness of and access to follow-up care in underserved populations should be considered. For example, focusing navigation programs toward individuals who have no record of follow-up care 180 days after an abnormal screening test result or those with more severely abnormal findings could reduce resource needs for programs while targeting those patients who may benefit most from the services.
There are several caveats that should be considered when interpreting the results of the current study. Although research costs were not included in the calculations, the overall efficiency of the PNRP navigation programs could have been reduced due to research data collection activities. These results may not represent the full range of costs or outcomes across the country or in settings or reimbursement models not represented within the PNRP sites. The study took a short-term perspective and did not include quality-adjusted LSY.[12, 23, 24] If new data emerge regarding the effectiveness of navigation, then future economic evaluations could extend the current analysis to address these longer-term outcomes.[12, 23]
The economic data from the current study are intended to be a source of evidence for decision-makers in health care delivery systems, public and private insurance plans, and government and nongovernmental organizations regarding the economic impact of patient navigation. However, decisions concerning the deployment of navigation will ultimately depend on the setting and needs of the population served, resources available, and public health priorities.
Funding was provided by the National Cancer Institute through its Center to Reduce Cancer Health Disparities contract 263-FQ-612391; by National Institutes of Health grants U01 CA116892 (Principal Investigator: Karen Freund), U01 CA117281 (Principal Investigator: Richard Roetzheim), U01CA116903 (Principal Investigator: Peter Raich), U01CA116937 (Principal Investigator: Steven Patierno), U01CA116924 (Principal Investigator: Kevin Fiscella), U01CA116885 (Principal Investigator: Donald Dudley), U01CA116875 (Principal Investigators: Steven Rosen and Elizabeth Calhoun), and U01CA116925 (Principal Investigator: Victoria Warren-Mears); and American Cancer Society grant SIRSG-05-253-01 (Principal Investigator: Electra Paskett).
Dr. Bensink was supported by grants from the National Cancer Institute, National Institutes of Health, and American Cancer Society to the Patient Navigation Research Program investigators and to Dr. Ramsey to complete this work. Dr. McKoy was supported by the Center to Reduce Cancer Health Disparities (grant 1K01CA134554-01). Dr. Seiber was supported by a grant from the American Cancer Society. Dr. Paskett was supported by grants from the National Cancer Institute, National Institutes of Health, and American Cancer Society. Dr. Mandelblatt's work was supported in part by National Cancer Institute grants U01 CA88283, U01CA152958, U01CA114593, and KO5CA96940.