Article, Toxicology

The impact of poison control centers on poisoning-related visits to EDs—United States, 2003

Original Contribution

The impact of Poison control centers on poisoning-related visits to EDsUnited States, 2003B

Eduard Zaloshnja PhDa,?, Ted Miller PhDa, Paul Jones PhDa, Toby Litovitz MDb,

Jeffrey Coben MDc, Claudia Steiner MD, MPHd, Monique Sheppard PhDa

aPacific Institute for Research and Evaluation, Calverton, MD 20705, USA

bNational Capital Poison Center, Washington, DC 20016, USA

cWest Virginia University Injury Control Research Center, P.O. Box 9151, Morgantown, WV 26506-9151, USA

dAgency for Healthcare Research and Quality, Office of Communications and Knowledge Transfer, Rockville, MD 20850, USA

Received 15 August 2007; revised 19 October 2007; accepted 21 October 2007

Abstract

Purpose: This study analyzes the association between center usage rates and the rates of nonadmitted visits to emergency departments (EDs) for poisoning.

Basic Procedures: With a log-normal regression model, we analyzed the association between the number of human exposure calls per hospitalized poisoning patient and the number of nonhospitalized ED visits. The data were from 14 states at county level.

Main Findings: A 1% higher poison control center (PCC) human exposure call rate for unintentional poisoning is associated, but not necessarily causally, with a 0.18% lower ED visit rate (P b .0001). If the observed association is causative, 15.5 PCC human poison exposure calls prevent one nonadmitted ED visit, yielding a $205 net cost saving and a benefit-cost ratio of 1.4. The savings ignore any reduction in hospital admissions.

Principal Conclusions: Increased PCC exposure calls appear to be associated with reduced ED use for unintentional poisoning and appear to reduce net medical spending.

(C) 2008

Introduction

Background

In a recent analysis, we showed that poison control center (PCC) use is associated with reduced hospital

? This work was funded in part by a Health Resources and Services Administration contract for technical assistance to the Federal PCC program and its grantees.

* Corresponding author. Tel.: +1 301 755 2700; fax: +1 301 755 2799.

E-mail address: [email protected] (E. Zaloshnja).

admissions for poisoning in rural areas [1]. In this study, we analyze the association between the PCC usage rate and the rate that poisoning patients are treated and released by urban and rural emergency departments (EDs).

More than 1.25 million patients, 440000 of them 0 to 19 years of age, were medically treated for poisoning in 2000 [2,3]. Toxicology professionals at PCCs provided free telephonic information and treatment advice regarding suspected Toxic exposures to a far greater number. They advised on 1.4 million exposures at 0 to 19 years of age alone according to Toxic Exposure Surveillance System data from the American Association of PCCs (AAPCC) [4,5].

0735-6757/$ – see front matter (C) 2008 doi:10.1016/j.ajem.2007.10.014

Importance

The value of promoting increased PCC use deserves careful scrutiny. Two studies provide credible evidence that 35% to 42% more poisonings would be treated in the ED and released in the absence of PCC consultation. Each study analyzes the change in use of other services when a PCC ceased operation. In early 1993, the Blodgett Regional PCC in Grand Rapids, Mich, stopped serving 2 of 3 area codes [6]. Outpatient claims to Blue Cross-Blue Shield of Michigan in the area codes that lost service rose to 35% [(969 – 715)/ 715]. By comparison, claim frequency was stable [(436 – 439)/439 = -0.7%) in the area code where service continued. When the Louisiana Regional PCC temporarily closed in November 1988, residents were instructed to call their local hospital. The hospitals in turn were encouraged to call an Alabama PCC [7]. Prior to closure, 25.0% of poisonings were medically treated. After closure, the Alabama center handled about one fourth of the prior Louisiana call volume. Among these cases, the medically treated poisoning rate was 74.4%. A lower bound estimate of the closing’s impact assumed the medically treated poisoning rate for the remaining cases as equal to the rate when the Louisiana center was operating [7]. This assumption seems more likely to underestimate than overestimate the actual usage rate. It implies that medically treated poisonings in Louisiana rose 42% following center closure [(35.6% of cases medically treated after closure – 25.0% treated during operation)/ 25.0%]. By comparison, the percentage medically treated was stable in neighboring Alabama.

Goals of this investigation

The posited sequence here is a simple one. Poison control center calls improve triage, potentially allowing home management of many exposures and avoiding ED visits. They also may avert ED visits when no treatment is needed. This paper estimates the association between PCC call volume and poisoning-related ED visit rates based on 2003 county-level data. Our analysis is correlational rather than a causative analysis based on tracking individual poisonings. We also conducted an economic analysis to estimate whether the reduction in ED visits associated with increased PCC call

volume was a cost-saving investment if it was causative.

Methods

Study design

We used a log-normal regression model to study the relationship between PCC use and poisoning-related ED visits that do not result in death or hospitalization. The dependent variable was the log of the county ED visit rate, defined as nonadmitted nonfatal ED visits for unintentional poisoning per

capita. The explanatory variable of interest was the log of PCC exposure calls from the county for unintentional poisoning per nonfatal hospital admission of county residents for uninten- tional poisoning. Poisoning-related hospitalizations thus served as a proxy for the underlying poisoning exposure level. The denominator includes poisoning-related hospitaliza- tions that were not managed by PCCs. This procedure implicitly assumes that PCCs do not reduce hospitalizations for poisoning. To the extent that they reduce it–and our companion article suggests they do in rural areas [1] –we will underestimate the association between PCC call volume and ED visits. We would have included poisoning-related fatalities in our proxy for the underlying poisoning exposure level, but the publicly accessible version of the National Vital Statistics System census of fatalities in the United States does not report county of residence for deaths occurring in counties with 100000 or fewer inhabitants.

Data collection and processing

The companion article describes the data sets we used. In brief, we analyzed 2003 State Inpatient and State ED Databases from the Healthcare Cost and Utilization Project and Toxic Exposure Surveillance System data on the number of human poison exposure calls per county from 14 states: Connecticut, Georgia, Indiana, Maine, Maryland, Massa- chusetts, Minnesota, Missouri, Nebraska, New Hampshire, South Carolina, Tennessee, Utah, and Vermont [8,9].

Using the same SAS algorithms as the companion article, we scanned the primary and secondary external cause (E) code and diagnosis fields of the ED and hospital discharge records for unintentional (or intent unknown) poisoning cases and eliminated readmissions.

Primary data analysis

The analysis used a log-normal regression model. When regressing ED visit rates against PCC call rates, we controlled for differences in admission policy and practices between US regions by creating dummy variables for Northeast/Mid-Atlantic, South, Midwest, and West. These variables capture differences related to hospital occupancy rates and admissions management practices of health insurers. Following the companion article, we created a dummy variable to distinguish rural counties from nonrural counties. County demographic characteristics controlled for were percentage of African-Americans, percentage of people 5 years of age and older who do not speak English very well, percentage of population living under the official level of poverty, and percentages of the population younger than 5 years and 5 to 19 years of age [10].

In sensitivity analysis, in order to eliminate the impact of outliers, we excluded counties with very low and high PCC human poison exposure call rates (the top and bottom 1%) and ran a median regression, which models the median

Table 1 ED visit and PCC call rates by state for rural and nonrural counties

State Population Poison-related ED visit rate (per thousand inhabitants)

PCC human poison exposure calls

PCC call rate (per hospital admission)

Nonrural

Rural

Nonrural

Rural

Nonrural

Rural

Nonrural

Rural

Conn

3 405565

3.56

28 202

9.06

Ga

7 695460

482564

3.06

4.52

70 926

3643

11.18

6.77

Ind

5 875711

204774

2.23

2.03

54 213

1746

9.86

12.67

Mass

6 324590

24507

4.05

7.43

42 843

263

7.06

12.81

Md

5 296486

2.76

44 576

9.67

Me

1 076575

198348

4.26

3.47

12 667

1678

9.34

8.01

Minn

4 324955

589239

1.81

1.40

40 682

4381

8.77

11.26

Mo

4 898695

696516

2.77

3.51

51 780

5988

7.71

6.28

Neb

1 319017

357853

1.81

1.59

14 834

2633

15.37

12.74

NH

1118551

117235

4.76

6.03

12 026

1172

12.48

.40

SC

3 985396

26616

3.11

2.63

26 789

113

7.25

4.16

Tenn

5 321137

368146

4.60

5.75

40 229

2346

6.49

5.26

Utah

2 104952

125335

3.53

3.83

38 846

1821

19.46

13.45

Vt

437512

171315

3.94

4.18

3861

1562

10.78

14.20

Total

53184602

3362448

3.17

3.39

482474

27346

9.19

8.25

instead of the mean. We also ran regressions without the Hospital readmission adjustment.

Economic analysis

The economic analysis used published estimates of poisoning treatment costs by place of treatment and of PCC costs per exposure call [1,3]. We inflated Medical costs to 2005 dollars using the consumer price index for hospital care. The costs to handle an exposure call were inflated using confidential salary data from nine PCCs while the portion of PCC staff time devoted to calls came from a single center [1].

Results

Descriptive statistics

In the nonrural counties of the 14 states analyzed, the rate of poisoning cases that required ED treatment but no hospitalization was 3.17 per thousand inhabitants. The ED visit rate ranged from 1.81 in Minnesota and Nebraska to 4.76 in New Hampshire (Table 1). The average PCC human poison exposure call rate was 9.2 per poisoning- related hospitalization, ranging from 6.5 in Tennessee to

19.5 in Utah.

Fig. 1 ED visit rates vs. PCC human poison exposures call rates per poison-related hospitalization by county (in log scale).

Table 2 Mean ED visit rate for unintentional poisoning per thousand population and mean PCC calls per unintentional poisoning hospitalization by quintile of the call rate distribution for 534 nonrural counties and 243 rural counties

Quintile Nonrural Rural

Mean calls per Mean ED hospitalization visit rate

Mean calls per hospitalization

Mean ED visit rate

1

4.64

4.05

3.77

3.69

2

7.08

3.97

5.23

3.83

3

9.06

3.29

8.97

3.19

4

11.95

3.23

12.65

2.55

5

26.65

2.82

31.21

2.12

In the rural counties, the average nonhospitalized ED visit rate was slightly higher, 3.39 per thousand inhabitants, ranging from 1.40 in Minnesota to 7.43 in Massachusetts. The average PCC human poison exposure call rate was slightly lower, 8.25 per poisoning-related hospitalization, ranging from 4.16 in South Carolina to 14.20 in Vermont.

Fig. 1 plots county level ED visit rates versus PCC human poison exposures call rates per hospitalization (on a log scale). Without controlling for other variables, ED visit rates visibly decline as the PCC call rate rises. With a few exceptions, the bottom of the distribution declines as PCC human poison exposure call rates increase. In addition, at around one third of the range, the highly dense center of the distribution starts trending down.

Table 2 shows the mean nonadmitted ED visit rate for poisoning and the PCC call rate per poisoning hospitalization by quintile of the county call rates. In both nonrural and rural areas, the visit rates decline as the call rates rise.

Regression results

Table 3 shows that an increasing PCC call rate is significantly, but not necessarily causally, associated with a decreasing ED visit rate. A 1% increase in the PCC human poison exposure call rate is associated with a 0.18% lower poisoning-related ED visit rate. That means an increase in calls across the 14 states of 5098 (0.01 x 509820 calls per year in the 14 states from Table 1) is associated with a Decrease in ED visits of 329.4 [0.0018 x (3.17 visits/1000 population x 53184.6 thousand nonrural residents + 3.39 visits/1000 population x 3362.45 thousand rural residents)]. Thus, evaluated at the means, if the observed association is causative, 15.5 PCC human poison exposure calls (5098.2/ 329.4) prevent 1 ED visit.

ED visit rates for unintentional poisoning are lower in counties with large populations of African Americans, people with poor English, or youth 5 to 19 years of age but higher in counties with a large proportion of children ages 0 to 4. Controlling for other factors, the ED visit rate for unintentional poisoning also differs by region. The rate is significantly much lower in the Midwest than elsewhere and

marginally higher in the Northeast/Mid-Atlantic than in the South. All variance inflation coefficients were small (b10), which indicates the absence of collinearity between regional dummy variables and the PCC call rates for unintentional human poison exposures (ie, call rates do not vary system- atically and significantly between regions).

Sensitivity analysis

Excluding the very low and high PCC call rates (the highest and lowest 1%), a 1% increase in the PCC call rates is associated, but not necessarily causally, with a 0.17% decrease (instead of 0.18%) in poisoning-related ED visit rates (P = .011). A median regression indicates that a 1% increase in the median PCC call rate is associated with a 0.16% decrease in the median of poisoning-related ED visit rates (P b.0001). Rerunning the models without adjusting for hospital readmission had no appreciable effect on the ED visit rate coefficients.

Economic analysis

It cost an estimated $34.94 to handle a human poison exposure call in 2005 [1]. Thus, $542 in spending on PCC services (15.5 human poison exposure calls x $34.94) would be needed to prevent one ED visit. Mean medical care payments for an ED visit for poisoning averaged $532 in 2000 ($745 inflated to 2005 dollars using the consumer price index for hospital care). Thus, $542 in PCC services is associated, but not necessarily causally, with a $745 decrease

Table 3 Log-normal regression results examining the association between ED visits for unintentional poisoning per thousand population and PCC calls per unintentional poisoning hospitalization

Variable

Parameter estimate

Variable impact

P N |t|

Intercept

2.3129

N/A

.000

Log (PCC call rate)

-0.1830

-0.18%

.000

Percent poverty

0.0003

0.001 ?

.000

% age 0-4

0.1380

0.458 ?

.000

% age 5-19

-0.0711

-0.236 ?

.000

% Afro-Americans

-0.0129

-0.043 ?

.000

% with poor English

-0.0663

-0.220 ?

.000

Rural counties

-0.2720

-31.3%

.000

West

0.1369

14.7%

.359

Midwest

-0.6547

-92.5%

.000

Northeast/Mid-Atlantic

0.1601

17.4%

.049

South

0

0.0%

0

Adjusted R2

.41

F10,766

54.04

P N F

b.0001

No. of observations

777

* Absolute change in mean ED visit rate for unintentional poisoning corresponding to 1 percentage point change in the variable.

in other medical spending, for a net saving of $203. If nationwide Specialist in Poison Information effort attributed to human exposure calls was not 83.2% but 80.3% or 86.3%, the cost per human exposure call would be $33.47 or $36.25 and the net savings $226 or $183, respectively.

From the health care system’s perspective, if the association is causal, this equates to a benefit-cost ratio of

1.4. Some spending would be necessary to increase PCC use. The break-even cost to obtain and handle a new call would be $48 ($745/15.5), $13 more than current mean cost per call handled. Mean call handling cost includes supervision, space, and other relatively fixed costs that do not rise unless large numbers of calls are added. Thus, the marginal cost of handling an additional call should be below the mean cost, which will improve the return on investment.

Discussion

Our analysis supports the findings from the 1990s. It suggests that a higher volume of calls to PCCs is potentially associated with lower poisoning-related visits to EDs. Although the results are statistically significant at a Confidence level above 99% and are robust, even if the observed association is causal, it might not hold for PCC call rates beyond the highest observed rate of 198 calls per hospitalization. The call rate eventually must reach a saturation point beyond which all remaining ED visits are necessary. Notably, the ED visit rates for unintentional poisoning for the 3 centers with call rates per hospitalization for unintentional poisoning above 100 were quite low, ranging from 0.17 to 0.44 per thousand population.

At the other extreme, only 4 counties had PCC call volumes below 2 calls per hospitalization for unintentional poisoning. The impact of completely eliminating a center might be larger than the impact of call volumes at these low levels.

The assumption that PCCs do not reduce hospitalizations for unintentional poisoning is a critical limitation that probably causes us to underestimate the reduction in ED visits associated with a higher PCC call rate. hospitalization rates fall significantly as PCC call volume rises in rural areas

[1] (although this association may not be causal), and the Grand Rapids experience suggests that eliminating PCC services might raise them significantly in urban areas as well. Adding the savings in rural admissions alone would raise the estimated return on investment to $1.75 per dollar currently spent on PCC services.

Calls that result in reduced ED visits are a mix of calls about nontoxic exposures and about toxic exposures amenable to home management, probably the least and most time-intensive exposure calls. Since calls that prevented ED visits are not identifiable, the average cost of these calls is unclear; it might be above or below the overall average.

The diagnosis coding in both hospital discharge and ED data does not very reliably identify the true intent of

poisonings–some intentional poisonings might be coded as unintentional. However, there is no evidence that this error is unevenly distributed among counties or between hospitaliza- tions and ED visits.

Lack of data forced us to omit health insurance coverage by county as an explanator in the regression models. Counties with less health insurance coverage may overuse EDs or delay medical care resulting in increased hospitaliza- tion for poisonings that would have not required admission if treated sooner.

Our rate data come from a convenience sample of 14 states, so our findings may not generalize to the United States as a whole. Nevertheless, those states house 29% of the total US population, and our findings clearly are representative for that large subset. Moreover, our rates miss any patients treated outside of those 14 states, notably patients in border counties who were treated in adjoining states. Cross-border leakages back and forth, however, at least partially cancel each other out. A final limitation is the exclusion of 26 rural counties that had no poisoning-related ED visits and/or hospitalizations. Several years of data would be necessary to calculate meaningful admission rates for those counties and would improve analytic precision in general. That suggests the need to replicate this analysis as more ED discharge data become available. With more years of data, the data set would include enough cases to support analysis by detailed age group and by type of exposure might improve our understanding of the observed association. Time series analysis of the effect of changing call volume in individual states also would be desirable.

Despite its limitations, our study establishes the existence of the hypothesized association between PCC utilization rates and ED visit rates for unintentional poisoning. It suggests restoring the AAPCC penetrance requirement for poison center certification. Supplemental case studies are desirable to assure that association mirrors a causative chain, detail any patient Outcome changes or reduction in hospital admissions that result from improved triage and immediate home treatment, and document the frequency and nature of unneeded ED visits for poisoning.

From the health care system’s perspective, ignoring any impact on hospital admissions, the break-even cost to obtain and handle a new call would be $48 or $13 more than the current mean cost per call handled. Our results indicate that $1.40 in ED costs is saved per dollar currently spent on PCC calls for unintentional poisoning exposures (and related outreach). The comparable estimate from an earlier study based on the Grand Rapids and Louisiana experience was $1.25, with another $4.90 in savings accruing from reduced admissions (and a total benefit-cost ratio of 6.15) [11]. An estimate for rural areas incorporat- ing our ED estimate and a recent estimate of the impact on rural admissions [1] is a combined savings of $8.30 per dollar spent on PCC unintentional exposure calls, again confirming the earlier study. These findings rebuke the

2007 Federal PCC funding cut that endangers these cost- saving medical services.

Acknowledgments

Our thanks to the 14 states that partner with the Healthcare Cost and Utilization Project for providing State ED and State Inpatient Databases to the project and to the PCCs in those states and the AAPCC for allowing us to use their deidentified data in this study. The analysis and opinions reported are those of the authors but not necessarily of the funding agency.

References

  1. Zaloshjna E, Miller TR, Jones P, et al. The impact of poison control centers on Rural hospitalization rates for poisoning. Pediatrics 2006;118:2094-100.
  2. Miller TR, Finkelstein AE, Zaloshjna E, Hendrie D. The cost of child and adolescent injuries and the savings from prevention. In: Liller K, editor. Injury prevention for children and adolescents: research, practice, and advocacy. Washington (DC): American Public Health Association; 2005. p. 15-64.
  3. Finkelstein EA, Corso PC, Miller TR, Fiebelkorn IA, Zaloshnja E, Lawrence BA. Incidence and Economic burden of Injuries in the United States, 2000. New York: Oxford University Press; 2006.
  4. Institute of Medicine- on Poison Prevention and control. Forging a poison prevention and control system. Washington (DC): National Academies Press; 2004.
  5. Rembert R. A profile of U.S. poison centers in 2002: A survey conducted by the American Association of Poison Control Centers; 2006. Available at: http://www.aapcc.org/2002_poison_center_survey_results. htm [Accessed March 29, 2007].
  6. Zech C. Letter to JR Maurer and JH Trestrail from Blue Cross and Blue Shield of Michigan, 8/12/93 in Poison Control Centers: is there an antidote for budget costs? Hearing before the Subcommittee on Human Resources and Intergovernmental Relations, Committee on Govern- ment Operations. 40th-41st ed. Washington (DC), 1994.
  7. King WD, Palmisano PA. Poison Control Centers: can their value be measured? South Med J 1991;84:722-6.
  8. HCUP SI Database Documentation. Healthcare cost and utilization project (HCUP). Agency for Healthcare Research and Quality. 3/13/

06. Available at: www.hcup-us.ahrq.gov/db/state/sidddbdocumenta- tion.jsp [Accessed March 2006].

  1. HCUP SEDD Database Documentation. Healthcare cost and utiliza- tion project (HCUP). Agency for Healthcare Research and Quality, Rockville, MD. 3/13/06. Available at: www.hcup-us.ahrq.gov/db/ state/sedddbdocumentation.jsp [Accessed March 2006].
  2. U.S. Census Bureau. Demographic profiles: 100-percent and sample data. Untied States Census Bureau Public Information Office. October 31, 2002. Available at: http://www.census.gov/Press-Release/www/

2002/demoprofiles.html [Accessed May 31, 2006].

  1. Miller TR, Lestina D. The costs of poisoning in the U.S. and the savings from poison control centers: a benefit-cost analysis. Ann Emerg Med 1997;29:239-45.