A revision to the United States national ALS registry’s algorithm to improve Case-Ascertainment

Abstract

Objective

To evaluate the impact of 1) updating the existing algorithm to improve case-finding sensitivity and 2) reclassifying the Registry’s diagnostic status nomenclature into four new categories (“confirmed ALS,” “likely ALS,” “undetermined ALS,” or “not ALS”) versus the current three (“definite ALS,” “possible ALS,” or “not ALS”) to be more inclusive and descriptive of cases and individuals.

Methods

A retrospective analysis of Registry data from 2011–2017 was conducted to follow “possible ALS” individuals over time to determine what qualifier caused them to convert, if at all and when, to Registry-eligible cases (i.e. “confirmed ALS” or “likely ALS”).

Results

In 2011, 720 individuals were classified by the Registry algorithm as having “possible ALS”. By 2017, 42% of these had converted to Registry-eligible ALS cases. Approximately 14% of those who were identified solely based on an ALS prescription drug never converted to Registry-eligible cases. This analysis indicates that “possible ALS” individuals with a single prescription for an ALS drug should be converted to Registry-eligible cases which would add between 300–500 cases per year on average.

Conclusions

The Registry’s existing algorithm likely results in the under-ascertainment of ALS cases. However, updating the algorithm with the inclusion of patients having been prescribed ALS-specific drugs, even with a single prescription, leads to improved epidemiologic estimates of ALS in the US. This and future algorithmic updates will help the Registry more accurately depict the true disease burden of ALS in the US.

Keywords:

• Amyotrophic lateral sclerosis
• registry
• case-ascertainment
• United States

Introduction

Amyotrophic lateral sclerosis (ALS), also known as known as Lou Gehrig’s disease, is a fatal neurodegenerative disease with significant neuromuscular involvement. The average life expectancy of the disease is 2 to 5 years from diagnosis (1–5). Up to approximately 32,000 people (9.9 cases/100,000 US population) are estimated to be living with ALS in the United States (US) (6). No cure for ALS has yet been identified, and the only two drugs currently approved by the US Food and Drug Administration (FDA) in the last 25+ years, Riluzole and Edaravone (introduced in 1995 and 2017, respectively), have been proven to slow disease, albeit modestly (7,8).

Since launching in 2010, the congressionally-mandated National ALS Registry (Registry) has helped to better describe the epidemiology (i.e. incidence, prevalence, mortality) of ALS in the US, examine risk factors such as environmental and occupational factors, and characterize the demographics of persons living with the disease (9–15). However, ALS, like most noncommunicable diseases, is not a nationally notifiable condition in the US, meaning that local and state health departments do not notify federal public health agencies about new cases. As such, the Registry developed and currently uses a novel two-pronged approach to identify US ALS cases (10). Briefly, the first approach uses a pilot-tested algorithm on large, federal administrative databases (e.g. Medicare), while the second approach uses a secure web-portal for persons with ALS to enroll in the Registry (10). Once cases are identified through the administrative databases and web-portal, data are annually merged, deduplicated, analyzed, and published.

In 2001, Congress passed landmark legislation to add ALS as a qualifying condition for automatic Medicare coverage and ended the waiting period for Social Security Disability Insurance (16). As such, up to 80% of ALS cases are found in the Registry’s administrative databases, specifically Medicare. Because of this, our paper focuses on the existing algorithm used to identify cases. While this algorithm has largely been effective in estimating the number of ALS cases in the US, the landscape of ALS has changed and evolved since the Registry’s launch in late 2010. Our primary objective here was to evaluate the impact on ascertainment of updating the Registry’s algorithm to ensure that Registry-eligible cases are not missed, and that case-capture is as timely and complete as possible. A secondary objective was to evaluate the impact of reclassifying the Registry’s diagnostic status nomenclature from three categories to four to be more inclusive and descriptive of cases and individuals.

Methods

The National ALS Registry’s existing case-ascertainment approach identifies prevalent cases through federal administrative databases (i.e. Medicare and Medicaid data from the Centers for Medicare and Medicaid Services [CMS] and military databases from the Veterans Heath Administration [VHA] and the Veterans Benefits Administration [VBA]). This is currently done by using the existing pilot-tested algorithm that includes elements such as the International Classification of Diseases (ICD) 9^th or 10^th revision encounter or diagnostic codes for ALS, frequency of visits to a neurologist, a death certificate listing ALS as a cause or contributing cause of death, and a prescription for Riluzole (Figure 1) (17).

Figure 1 Original and proposed algorithms for defining cases in the National ALS Registry.

Once individuals are identified, the Registry currently classifies them as either “definite ALS” cases, “possible ALS” individuals, or “not ALS” individuals, and only “definite ALS” cases are Registry-eligible (Table 1, Figure 1). Registry-eligible cases are the only ones used as numerator cases to determine prevalence rates. Similarly, for purposes of this paper, the term “case” only refers to Registry-eligible cases; all else are referred to as individuals (e.g. “possible ALS” individuals). “Definite ALS” cases are ones who meet the Registry case criteria and are Registry-eligible. “Possible ALS” individuals are ones in which there is not enough information to determine status in either direction (i.e. either “definite ALS” or “not ALS”), and they are not yet Registry-eligible. Records of these “possible ALS” individuals are set aside until subsequent information is received so that a status determination can be made later. “Not ALS” individuals are ones who clearly do not meet the case definition and are not Registry-eligible. Finally, Registry-eligible cases remain in the Registry until confirmed as deceased through the National Death Index (NDI) (18). Cross-referencing the living ALS cases from the Registry with NDI allows the Registry to maintain a current listing of cumulative prevalent cases.

Table 1 Current and proposed classifications of individuals in the national ALS registry.

Registry-Eligible Cases^a	Current^b	Proposed^c
Yes	Definte ALS	Confirmed ALS^a
		Likely ALS^a
No	Possible ALS	Undetermined ALS
	Not ALS	Not ALS

^aOnly Registry-eligible cases will be added to the numerator in order to determine rates.

^bFor years 2011–2017.

^cFor years 2018 and forward.

In 2021, the National ALS Registry convened an expert panel of neurologists to discuss the current therapeutic landscape and recommend revisions to the existing algorithm to improve case-ascertainment and epidemiological estimates nationally for the disease. Two main recommendations came from the panel. The first was to examine “possible ALS” individuals with only a prescription of Riluzole to determine how, when, or if these individuals convert to Registry-eligible cases. This recommendation was made because Riluzole is typically only used for ALS; although off-label use exists, it is minimal at most. The second recommendation was to reclassify Registry’s diagnostic status nomenclature into four new categories versus the current three to be more inclusive and descriptive of cases and individuals (Table 1).

For the first panel recommendation, a retrospective analysis of Registry data was conducted to follow “possible ALS” individuals across years (2011–2017) to determine what qualifier (e.g. ICD code, ALS prescription drug) caused them to convert to Registry-eligible cases, if at all, and when (Figure 2). Individuals who did not convert were examined more closely to better understand why.

Figure 2 Flow chart shows the filtering of Possible ALS individuals in current algorithm. ^aD1 definition: 1) an encounter coded for ALS (International Classification of Diseases, Ninth Revision ICD-10 G12.21) in 1 or more years in the same source, or 2) a death certificate listing ALS as a cause of death, or 3) a prescription for Riluzole.

Regarding the second panel recommendation, modifying the existing classifications (“definite ALS”, “possible ALS”, or “not ALS”) to the new classifications (“confirmed ALS”, “likely ALS”, “undetermined ALS”, or “not ALS”) better qualifies actual case and individual status (Table 1, Figure 1). Under the new classification, “confirmed ALS” and “likely ALS” cases are Registry-eligible, whereas “confirmed ALS” cases are ones where ALS encounter codes or a death certificate is available, and “likely ALS” cases are ones where a single prescription for an ALS therapeutic such as Riluzule or Edaravone is available. These “confirmed” and “likely” ALS cases will be entered as numerator cases when estimating ALS prevalence rates. Conversely, if insufficient or no data are available, such individuals will be categorized as “undetermined ALS” or “not ALS”, respectively (Table 1). Individuals who are classified with “undetermined ALS” could eventually convert to “likely ALS” or “confirmed ALS” cases when or if additional data are available and analyzed.

Results

In 2011, 720 “possible ALS” individuals were identified by the Registry with only a single Riluzole prescription. Of these 720 individuals, 42% (n = 305) converted to a Registry-eligible case (“likely ALS”) by the end of 2017 and 18% (n = 131) converted to a “confirmed ALS” case when a death certificate was received (Table 2). In 2016, the number of “possible ALS” individuals with only a prescription of Riluzole was higher (n = 964) but only 22% (n = 215) had converted by the end of 2017 (Table 2). Of these 215 cases that converted to Registry-eligible cases from 2016 to 2017, over 90% converted with either an encounter coded as ALS or a death certificate (“confirmed case”).

Table 2 ALS patients with a Riluzole prescription in administration databases -- Current Algorithm.

	Possible ALS individuals
	2011		2012		2013		2014		2015		2016		2017
	n	%	n	%	n	%	n	%	n	%	n	%	n
Number of possible^a	720		558		311		16		1,079		964		411
Number converted to Confirmed case by 2017	305	42.4	213	38.2	96	30.9	6	37.5	395	36.6	215	22.3	N/A
Number converted to Confirmed case from D1^b definition by 2017	186	25.8	123	22.0	19	6.1	6	37.5	239	22.2	194	20.1	N/A
Number converted to Confirmed case with Riluzole and death certificate by 2017	131	18.2	67	12.0	5	1.6	0	–	91	8.4	77	8.0	N/A

^aPossible ALS individuals with only a Riluzole prescription.

^bD1 - ALS confirmed case if any two of the following: 1) an encounter coded for ALS (International Classification of Diseases, Ninth Revision ICD-10 G12.21) in 1 or more years in the same source, or 2)a death certificate listing ALS as a cause of death, or 3) a prescription for Riluzole.

Under the existing algorithm, in 2011 there were 12,187 “definite ALS” cases in the Registry (Table 3). Using the newly proposed algorithm, we would be retaining all 720 formerly known as “possible ALS” individuals as “likely ALS” cases and count them toward ALS Registry-eligible cases, due to their qualifier – a single prescription. This modification would yield a total of 12,907 Registry-eligible cases. Following this pattern, an increase of “likely ALS” cases is seen for all subsequent years (Table 3). For 2017, which is the most current year of prevalence data available, an additional 1,584 ALS Registry-eligible cases would have been added for a total of 19,384 (Table 3). Over time, some of the “likely ALS” cases could become “confirmed ALS” cases as the Registry obtains more validating information (e.g. medical encounter code, NDI death certificate stating ALS as a cause of death). Nonetheless, by introducing this new classification, the total ALS cases in each year will be inclusive of those who are eligible by a single qualifier alone, reducing the likelihood of undercounting the persons with ALS living in the US.

Table 3 Number of ALS cases^a using the newly proposed algorithm and classification system for the National ALS Registry.

	Years
ALS Classification^b	2011	2012	2013	2014	2015	2016	2017
Confirmed	12,187	14,713	15,908	15,927	16,583	16,424	17,800
Likely^c	720	844	650	54	1,226	1,364	1,584
Total ALS cases	12,907	15,557	16,558	15,981	17,809	17,788	19,384

^aCumulative by year.

^bNew classifications: Confirmed and Likely ALS cases (i.e. Registry-eligible cases) equals "definite ALS" cases under existing classification.

^cLikely ALS cases are the newly added cases under the new algorithm.

Discussion

The congressionally mandated National ALS Registry is the only public health surveillance system in the US that examines the epidemiology (e.g. prevalence) of the disease at the national level. Because ALS is non-notifiable in the US, it is imperative that the sensitivity of the Registry’s case-finding algorithm is sufficient to ensure that Registry-eligible cases are not missed, and case-capture is as timely and complete as possible. Increased sensitivity allows the Registry to better describe the public health burden of ALS nationally.

The recent capture-recapture publication by Nelson et. al. showed that the Registry may be missing up to approximately 44% of cases in the US. Many of these missing cases are likely explained by people with ALS not utilizing federal health insurance, but instead using private insurance companies (e.g. HMOs, PPOs) to which we do not have database access (19). When extrapolated to the most recent prevalence estimates available for 2017, the Registry may be missing as many as 14,000 cases (6). To improve ascertainment of these missing cases, a revision in the algorithm is necessary to help account for the under-ascertainment of cases.

The inclusion of a single prescription of Riluzole and/or Edaravone will improve sensitivity in the identification of more ALS cases as Registry-eligible (“likely ALS”). This will improve the national ALS estimates as identified by the Registry. Moreover, this will result in a re-analysis of the existing capture-recapture estimate for future prevalent years (≥2018). This new capture-recapture methodology is currently being undertaken by the Registry and results have not been finalized.

The re-classification of the Registry’s diagnostic status nomenclature supported by this analysis avoids confusion with the current El Escorial naming scheme, will allow for a more comprehensive and inclusionary approach to those with a single prescription of either Riluzole or Edaravone now considered “likely ALS” (20). Our analyses of the Medicare Part D Prescribers database shows that Riluzole is primarily being used for the treatment of ALS (21). As new ALS therapeutics are approved and added to formularies for treatments, the Registry will continue to evaluate each one on a case-by-case basis. If it is determined these therapeutics will primarily be used for ALS, then it is our intent to add them as a criterion for a Registry-eligible case (i.e. “confirmed” or “likely” ALS) as well as periodically evaluate the algorithm and make updates, when necessary.

Moreover, as the ALS treatment landscape evolves, the Registry will continue to make every effort to improve case-ascertainment methods by adding new data sources, when possible. This can be achieved by evaluating sources of ALS cases such as data from new state-based registries or nonprofit patient organizations and collaborating with private insurance companies. The Registry is committed to providing the public with the most reliable estimates of ALS disease burden in the US.

Conclusions

The ascertainment of ALS cases in the US continues to be challenging, specifically when ALS is non-notifiable and when patients receive their care from private insurance carriers, currently not accessible by the National ALS Registry. The data analyses conducted by the Registry support the adjusting of the case-finding algorithm and reclassifying cases into more descriptive categories, beginning with calendar year 2018 data. We feel this will improve overall case-ascertainment resulting in better epidemiological estimates going forward. The Registry is committed to adapting to the ever-changing epidemiologic disease landscape and implementing measures to improve and increase case-ascertainment to better describe the public health burden of ALS nationally.

Disclaimer

The findings and conclusions in this report are those of the authors and do not necessarily represent the official positions of the Agency for Toxic Substances and Disease Registry, the Centers for Disease Control and Prevention, and/or the US Department of HHS.

Acknowledgements

The authors are grateful to those living with ALS who give their valuable time to contribute important health data to researchers. Without their help, these findings, and countless others, would not be possible.

Declaration of interest

The CDC/ATSDR authors have no declarations of interest. Dr. Berry, Dr. Brooks, Dr. Oskarsson, and Dr. Goutman have no declarations of interest linked to the study.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This work was supported by the Centers for Disease Control and Prevention (CDC)/Agency for Toxic Substances and Disease Registry (ATSDR). The findings and conclusions in this report are those of the authors and do not necessarily represent the views of CDC/ATSDR.