Communicating Health Economic Evidence: From Model Outputs to Policy Briefs

อ่านฉบับภาษาไทย (Thai version)

Abstract

A province's health fund must decide whether to pay for C, a preventive treatment. The analyst's draft said C was good value at 150,000 THB per quality-adjusted life year (QALY), which is a year of life weighted by its quality. It added "60% sure: pay for it" and answered none of the board's questions. This article rewrites it as a policy brief. The decision and deadline come first, then four options with the same lines. Value: 10,000 THB under what the board will pay for a QALY, 160,000 THB. The yearly bill: 100, 200 and 300 million THB against a yearly budget of 200 million. How sure: 599 of 1,000 reruns with uncertain inputs redrawn favour C, which is not the chance a patient benefits. Who gains: town-first starts give the town 925 QALYs and cost rural districts 675 QALYs. A drug price cut from 60,000 to 40,000 THB a year improves every line. It ends with conditions and a review date.

Summary card: the title, the key idea that a brief opens with the board's decision, its deadline and its options and gives every option the same four lines, and five takeaways for treatment C in one province: real options, from the list price and a drug price cut to 40,000 THB to waiting a year; 1.00 QALY for 150,000 THB more, 10,000 THB under the threshold of 160,000 THB; a yearly bill of 100, 200 and 300 million THB against a budget of 200 million; 599 of 1,000 reruns favour C, a statement about the model's inputs, while on C 9.55 fewer people in every 100 have a complication; town first gives the town +925 QALYs and leaves rural districts 675 QALYs worse off. — Visual summary · ภาพสรุป

A verdict the board could not use

A provincial health fund pays for care across a province where 10,000 adults are eligible for C, a new preventive treatment for a chronic condition. The fund's board meets in 4 weeks to set next year's budget. It has 20 minutes for C on the agenda and will read at most 2 pages beforehand.

Per patient, C costs 150,000 THB (Thai baht) more than usual care. It adds health equal to one year in full health, once every year lived is weighted by how well it is lived: 1.00 QALY. A QALY, a quality-adjusted life year, is a year of life weighted by its quality, from 1 for full health to 0 for death. Over 10 years, a state valued at 0.70 counts as 7.0 QALYs, so restoring full health gains 3.0; the full lesson is in QALY, DALY and the Utilities Behind Them: What Is Actually Being Measured.

Divide C's extra cost by its extra health, both measured against usual care, the next best option. That gives 150,000 THB for 1.00 extra QALY, or 150,000 THB per extra QALY. The incremental cost-effectiveness ratio (ICER) is this extra cost divided by the extra health gained against the next best option. The fund sets it against a threshold of 160,000, the most its board is willing to pay for one extra QALY.

The ICER article also covers the rule that removes an option before any ratio is taken. In the series' shared four-option example, separate from this article's treatment C, one option costs 270,000 THB for 5.90 QALYs and another costs 250,000 THB for 6.00 QALYs. The first costs 20,000 THB more and gives 0.10 QALY less, so it is ruled out; this rule is called dominance. Here usual care is the comparator; the full lesson is in ICER, Dominance and the Cost-Effectiveness Plane: Reading the Thai Threshold of 160,000 Baht per QALY.

The fund's analyst sent the board a first draft of one paragraph: "C is cost-effective. Its ICER of 150,000 THB per QALY is under the threshold, and the model gives a 60% probability that C is cost-effective. We recommend adoption."

Every number in it was right, yet the board could not use it. The finance member asked what C would cost the fund each year and whether the budget could take it. A doctor read "60%" as "6 in 10 patients benefit". A member from a rural district asked who would get C first, and nobody could say what the fund could do besides a plain yes or no.

The draft rests on a common belief: that a brief is the model's verdict in short form, cost-effective, so adopt. A brief is written for the people who decide. It starts from their decision and their options, then gives each option the same few lines: value, the yearly bill, how sure the evidence is and who gains [1]. This article rewrites the analyst's paragraph that way, with the province's numbers.

Start with the decision

Before any model number, the brief says what the board must decide. The board decides whether the fund pays for C from next year, at what price and for whom, at its budget meeting in 4 weeks. If it does nothing, the 10,000 eligible adults stay on usual care and the fund spends nothing new on C.

Then it lists the options, because the board chooses between actions, not between a yes and a no [2]. There are four options here. The fund can pay the list price, or pay only if the maker lowers it. It can pay a lower price and start rural districts at the town's pace, or wait a year and gather evidence.

Each option gets the same lines, so the board can compare them side by side. A short document built this way has a name. A policy brief is a few pages for the people who must choose: it sets out a decision, its options and what each is likely to bring, including what is still uncertain [1]. The model's methods go in an appendix, where an analyst can check them, not in the opening.

Value: the size before the ratio

A ratio on its own hides size. The same 150,000 THB per QALY could come from a tiny gain at a tiny cost or a large gain at a large cost. So the value line gives both parts first: per patient, C adds 1.00 QALY and costs 150,000 THB more over its three-year course. Only then does it give the ratio, 150,000 THB per QALY, and the threshold, 160,000 THB.

The board also needs one number for how far under the threshold C sits. At the threshold each QALY is worth 160,000 THB, so C's 1.00 QALY is worth 160,000 THB. Take away its extra cost of 150,000 THB, and 160,000 − 150,000 = 10,000 THB of value per patient is left over.

This margin has a name: net monetary benefit (NMB) is health valued at the threshold minus cost. The incremental NMB (INMB) is that margin against usual care: 1.00 extra QALY for 150,000 THB more gives 10,000 THB; the full lesson is in ICER, Dominance and the Cost-Effectiveness Plane: Reading the Thai Threshold of 160,000 Baht per QALY.

In the brief the value line reads: "Per patient, C adds 1.00 QALY for 150,000 THB more. At the 160,000 THB threshold that is 10,000 THB more value than cost." It says nothing about how many patients there are, what the fund pays each year or who goes first.

The yearly bill: the same treatment, counted per year

Read first on this site:

The finance member's question needs a different count: everyone on C, not one patient. The province's clinics can start 2,000 people a year, and each stays on C for three years. So 2,000, 4,000 and 6,000 people are on C in years 1, 2 and 3 of the plan. Each year on C costs the fund 60,000 THB for the drug and saves 10,000 THB of complication care: 50,000 THB more than usual care.

So the fund pays 2,000 × 50,000 = 100 million THB in year 1, 200 million in year 2 and 300 million in year 3. That is 600 million THB over three years. Its budget for new treatments is 200 million THB a year, so year 3 is 100 million THB over [5].

The bill and the value line use the same 50,000 THB a year, but count it differently. The value line adds one patient's three years, 3 × 50,000 = 150,000 THB; the bill adds everyone on C in one year.

The two line up only by assumption: the brief takes the ICER's cost as the same 50,000 THB a year, with nothing taken off for money paid later. In a lifetime model they rarely line up, so the brief states this assumption and says which number is per patient and which is per year.

The same money buys health. Count the course's 1.00 QALY evenly over its three years, so 3 person-years on C buy one QALY (a person-year is one person treated for one year). The plan's 2,000 + 4,000 + 6,000 = 12,000 person-years then buy 12,000 / 3 = 4,000 QALYs, the same as 600 million / 150,000. People who start in years 2 and 3 finish their course after the plan ends, so only part of it is counted.

That money also has to come from somewhere. The threshold is what the board chooses to pay for a QALY, while the cost per QALY of the care pushed out is a fact of the fund that nobody knows exactly. Suppose it is also 160,000 THB: the 600 million THB would have bought 3,750 QALYs there, health pushed out because the money went to C. The net gain is 4,000 − 3,750 = 250 QALYs.

In the brief the bill line reads: "100, 200 and 300 million THB in years 1 to 3, against a budget of 200 million a year. Year 3 is 100 million over."

R code annual_budget.R

# The fund's yearly bill for C at the list price, the health it buys and pushes out, and who gains.
on_c <- 2000 * 1:3                       # people on C in years 1, 2 and 3 (a three-year course)
drug <- 60000; offset <- 10000           # THB a person-year: the drug; complication care it saves
bill <- on_c * (drug - offset) / 1e6     # million THB
cat("People on C:", on_c, "| person-years:", sum(on_c), "| extra THB each:", drug - offset, "\n")
cat("Bill (million THB):", bill, "| total:", sum(bill), "| year 3 over 200:", bill[3] - 200, "\n")
k <- 160000                              # THB per QALY: the threshold
gained <- sum(on_c) / 3; lost <- sum(bill) * 1e6 / k     # 3 person-years buy 1 QALY
cat("QALYs bought on C:", gained, "| pushed out elsewhere:", lost, "| net:", gained - lost, "\n")
starts <- c(town = 1400, rural = 600)    # starts a year, town first (5,000 eligible in each)
py <- starts * (3 + 2 + 1)               # person-years: year-1 starts count 3 years, then 2, then 1
cat("Person-years:", py, "| QALYs:", py / 3, "| pushed out each:", lost / 2,
    "| net:", py / 3 - lost / 2, "\n")

Output of the run annual_budget.log

People on C: 2000 4000 6000 | person-years: 12000 | extra THB each: 50000
Bill (million THB): 100 200 300 | total: 600 | year 3 over 200: 100
QALYs bought on C: 4000 | pushed out elsewhere: 3750 | net: 250
Person-years: 8400 3600 | QALYs: 2800 1200 | pushed out each: 1875 | net: 925 -675

The fund's yearly bill at the list price, the year-3 overrun, and the QALYs the money buys on C and pushes out elsewhere.

The sliders start at year 3 of the case, with 6,000 of the 10,000 eligible on C at 50,000 THB a person-year. So uptake, the share of the eligible on C, starts at 60, and the year-3 bill is 300 million THB. The value tile uses one patient's three-year course: 1.00 QALY for 150,000 THB more is worth 10,000 THB more than its cost at the 160,000 THB threshold. Raise the uptake and only the bill tile grows; lower the threshold below 150,000 THB and only the value tile turns red, so the brief keeps the two on separate lines.

How sure: what 0.60 means

Read first on this site:

Every input of the model is uncertain: the trial behind C was small, and the costs and quality weights come from samples too. So the analysts reran the model 1,000 times, each time drawing every uncertain input afresh from the range its evidence allows. Across the reruns C's net benefit averaged 10,000 THB but swung widely, with a standard deviation of 40,000 THB: the typical distance of a rerun from that average. Probabilistic sensitivity analysis (PSA) reruns the model many times with every uncertain input drawn from its distribution, so the result is a share of runs rather than one number; the full lesson is in Probabilistic Sensitivity Analysis: From Parameter Distributions to CEAC.

So the average sits 10,000 / 40,000 = 0.25 standard deviations above zero. The reruns spread in a bell shape around their average, and with the average that close to zero, about 599 of the 1,000 land above it: a share of 0.60 [4]. That is the draft's "60% probability".

It is a statement about the model. Given this evidence, this model and the 160,000 THB threshold, C comes out worth its price in 6 of every 10 reruns.

It is not the chance that a patient benefits. In the trial behind C, 13 of 60 people on usual care had a complication during follow-up, a risk of 21.67%, against 8 of 66 on C, a risk of 12.12%. That is 21.67 − 12.12 = 9.55 fewer complications for every 100 people treated. Most people on C would not have had a complication anyway, so "6 in 10 patients benefit" is wrong by a wide margin.

Nor is it the chance that the model is right. The reruns vary only the inputs inside the model. If the trial was biased or the model leaves out a cost, every rerun carries the same error, and more reruns cannot remove it.

The board also needs to know how far the trial itself can be trusted. C's effect rests on one small trial with 8 + 13 = 21 complications. Its risk ratio, 12.12 / 21.67 = 0.56, has a 95% confidence interval of 0.25 to 1.26.

How far a body of evidence can be trusted for one outcome is its certainty. One small trial whose interval spans both a large benefit and harm leaves the true effect unclear. So the certainty is rated low; the rating rules are taught in Beyond the Evidence Pyramid: GRADE Certainty and Choosing Effectiveness Inputs for Economic Models.

In the brief the how-sure line reads: "In 599 of 1,000 reruns of the model, C was worth its price at 160,000 THB per QALY. This describes the uncertainty in our inputs, not patients: on C, 9.55 fewer people in every 100 have a complication. That effect comes from one small trial, so the evidence is of low certainty."

Repeat the count at another threshold and the share changes. At 200,000 THB per QALY the net benefit is 200,000 − 150,000 = 50,000 THB, 1.25 standard deviations above zero, so about 894 of 1,000 reruns favour C. A cost-effectiveness acceptability curve (CEAC) shows, at each threshold, the share of simulation runs in which a treatment is the best buy [4]. At 160,000 THB per QALY it reads 0.60, because about 599 of 1,000 runs favour C; the full lesson is in Probabilistic Sensitivity Analysis: From Parameter Distributions to CEAC.

Read the curve at the threshold the payer actually uses, here 160,000 THB, and as a statement about the model, never about patients.

R code value_summary.R

# One patient on C against usual care: the value line of the brief, and what 0.60 means.
d_cost <- 150000; d_qaly <- 1.00; k <- c(160000, 200000)  # THB more; QALYs more; thresholds
nb <- k * d_qaly - d_cost                                  # net benefit at each threshold, THB
cat("Cost per extra QALY:", d_cost / d_qaly, "THB | net benefit at each threshold:", nb, "THB\n")
spread <- 40000                            # THB: standard deviation of net benefit across reruns
z <- nb / spread; share <- pnorm(z)        # spreads above zero; share of reruns above zero
cat("Spreads above zero:", z, "| reruns favouring C of 1000:", round(1000 * share),
    "| share:", sprintf("%.2f", share), "\n")
risk <- round(100 * c(usual = 13 / 60, on_c = 8 / 66), 2)   # % with a complication in the trial
cat("Trial risk (%):", risk, "| fewer per 100 treated:", risk[1] - risk[2], "\n")
quality <- 0.70; years <- 10               # QALY recap: 10 years in a state worth 0.70
cat("QALYs:", sprintf("%.1f", c(quality * years, (1 - quality) * years)), "\n")

Output of the run value_summary.log

Cost per extra QALY: 150000 THB | net benefit at each threshold: 10000 50000 THB
Spreads above zero: 0.25 1.25 | reruns favouring C of 1000: 599 894 | share: 0.60 0.89
Trial risk (%): 21.67 12.12 | fewer per 100 treated: 9.55
QALYs: 7.0 3.0

One patient on C: the price per extra QALY, the net benefit at two thresholds, the share of 1,000 reruns that favour C, and the trial's complication risks.

R code evidence_certainty.R

# The trial behind C: its two risks, the risk ratio and its 95% confidence interval.
events <- c(on_c = 8, usual = 13); n <- c(on_c = 66, usual = 60)   # complications; patients
risk <- events / n                                   # risks over the trial's follow-up
rr <- risk[["on_c"]] / risk[["usual"]]               # risk ratio, C against usual care
se <- sqrt(sum(1 / events - 1 / n))                  # standard error of log(risk ratio)
ci <- exp(log(rr) + c(-1, 1) * qnorm(0.975) * se)    # 95% confidence interval
cat("Complications:", sum(events), "| risks (%):", sprintf("%.2f", 100 * risk), "\n")
cat("Risk ratio:", sprintf("%.2f", rr), "| 95% CI:", sprintf("%.2f", ci), "\n")

Output of the run evidence_certainty.log

Complications: 21 | risks (%): 12.12 21.67
Risk ratio: 0.56 | 95% CI: 0.25 1.26

The trial behind C: complications, the two risks, and the risk ratio with its 95% confidence interval.

R code dominance_example.R

# The series' shared four-option example: one option costs more and gives less health than another.
cost <- c(first = 270000, second = 250000)   # THB
qaly <- c(first = 5.90, second = 6.00)       # QALYs
cat("Costs (THB):", cost, "| QALYs:", sprintf("%.2f", qaly), "\n")
cat("First against second:", cost[["first"]] - cost[["second"]], "THB more,",
    sprintf("%.2f", qaly[["second"]] - qaly[["first"]]), "QALY less\n")

Output of the run dominance_example.log

Costs (THB): 270000 250000 | QALYs: 5.90 6.00
First against second: 20000 THB more, 0.10 QALY less

The series' shared example from the opening: two options' costs and QALYs, and how much more the first costs for how much less health.

Who gains, and who waits

Half of the 10,000 eligible adults live in the main town near the provincial hospital and half in rural districts. If clinics start people where they are ready, 1,400 of each year's 2,000 starts are in town and 600 are in rural districts.

Counting 3, 2 and 1 years for people who start in years 1, 2 and 3, the town's starts add up to 1,400 × 6 = 8,400 person-years on C and the rural starts to 600 × 6 = 3,600. At 3 person-years per QALY, the town gains 8,400 / 3 = 2,800 QALYs and the rural districts 1,200. On the same supposition, the 3,750 QALYs pushed out are care that everyone uses, so each half gives up 1,875.

The town nets 2,800 − 1,875 = 925 QALYs, and the rural districts 1,200 − 1,875, a loss of 675 QALYs. They give up as much other care as the town but get far fewer people on C. The total, 925 − 675 = 250 QALYs, matches the bill line and hides that one group ends up worse off.

In the brief the fairness line reads: "Starting where clinics are ready gives the town 925 QALYs. It leaves rural districts 675 QALYs worse off." A brief names who gains, who pays and who waits, even when the total looks good.

Facts more evidence can settle, choices the board must make

The certainty line and the fairness line raise different kinds of doubt. Whether C really prevents 9.55 complications for every 100 people treated is a question of fact. A larger trial, or a register of the first people treated in the province, can narrow it, so money spent on evidence can shrink this doubt.

Whether rural districts should start C at the same pace as the town is not a question of fact. No trial can tell the board how much a rural QALY should count; the board must choose and say why. A brief keeps the two apart, so that a value choice is not dressed up as a number and a gap in the evidence is not settled by opinion.

Options, not a verdict

The draft offered one action. The brief gives the board four and runs each through the same lines [2]. Option 1 pays the list price and starts people where clinics are ready. That is everything above: 150,000 THB per QALY, a bill of 100, 200 and 300 million THB, 599 of 1,000 reruns, and rural districts 675 QALYs worse off.

Option 2 pays only if the maker cuts the drug from 60,000 to 40,000 THB a year. A lower price moves every line at once, so the worked example below follows it through.

What a drug price of 40,000 THB does to every line

Option 2 keeps everything else from option 1. Each patient still gains 1.00 QALY, 10,000 THB of complication care is still saved each year, and the start is still town first.

Extra cost per year on C

\[ 40000-10000=30000 \]

The drug now costs 40,000 THB a year and still saves 10,000 THB of complication care. So each year on C costs the fund 30,000 THB more than usual care.
The value line

\[ 3\times 30000=90000,\quad 160000-90000=70000 \]

Over the three-year course one patient costs 90,000 THB more for the same 1.00 QALY. So the price per extra QALY is 90,000 THB, and the net benefit at the 160,000 THB threshold is 70,000 THB.
The bill line

In years 1, 2 and 3, the 2,000, 4,000 and 6,000 people on C cost 30,000 THB each. That gives 60, 120 and 180 million THB, every year under the budget of 200 million THB.
The how-sure line

If the standard deviation across reruns stays at 40,000 THB, the average of 70,000 THB sits 70,000 / 40,000 = 1.75 standard deviations above zero, so about 960 of 1,000 reruns favour C. That is up from 599 at the list price.
Health overall

\[ 60+120+180=360,\quad 4000-2250=1750 \]

The three years cost 360 million THB. That buys the same 4,000 QALYs on C and pushes out 360 million / 160,000 = 2,250 QALYs elsewhere. The net gain is 1,750 QALYs.
The fairness line

\[ 2800-1125=1675,\quad 1200-1125=75 \]

Each half now gives up 2,250 / 2 = 1,125 QALYs. With the same town-first start the town nets 1,675 QALYs and the rural districts 75 QALYs: no group is worse off, but the gap between them stays wide.

Result: At 40,000 THB every line improves. The price per QALY is 90,000 THB, and the year-3 bill of 180 million fits the budget. Reruns favouring C rise to 960 of 1,000, and the net gain is 1,750 QALYs against 250 QALYs at the list price. The price is the strongest lever the board holds, so the brief sets the negotiated price beside the list price.

An equal start, or waiting a year

Option 3 takes the negotiated price and starts 1,000 people a year in each half. A mobile clinic costing 32 million THB, bought in year 1, reaches rural patients, so the bill becomes 92, 120 and 180 million THB. The clinic adds cost but no QALYs: 360 + 32 = 392 million THB buys the same 4,000 QALYs, or 392 million / 4,000 = 98,000 THB per QALY. The net benefit per patient is 160,000 − 98,000 = 62,000 THB, 1.55 standard deviations above zero, so about 939 of 1,000 reruns favour C.

Each half now has 1,000 × 6 = 6,000 person-years and gains 2,000 QALYs. The 392 million THB pushes out 2,450 QALYs, 1,225 from each half, so each nets 2,000 − 1,225 = 775 QALYs. The total falls to 1,550 QALYs, 200 QALYs below option 2 at the same drug price.

Those 200 QALYs are what the province gives up for an equal start. It is a value choice that the brief states but does not make.

Option 4 waits a year and pays for a register of the first patients or a larger trial. Nothing new is spent on C and nobody starts it, so the 2,000 people who would have started in year 1 wait. Waiting is worth it only if the evidence it buys could change the board's choice.

⤢ click to enlarge

Read across each row: green cells meet the board's line (under the 160,000 THB threshold, within the 200 million THB budget, health gained) and orange cells miss it. Only option 1 has orange cells, its year-3 bill of 300 million and the rural districts' loss of 675 QALYs. Compare rows 2 and 3: the same drug price and year-3 bill, but the clinic raises the cost per QALY to 98,000 THB and the equal start gives each half 775 QALYs instead of 1,675 and 75. Net QALYs are QALYs gained minus QALYs pushed out, on the 160,000 THB supposition.

R code brief_options.R

# Three ways to pay for C, side by side: price per QALY, yearly bill, net health, who gains.
k <- 160000; offset <- 10000; spread <- 40000       # THB per QALY; THB a person-year; THB
drug <- c(60000, 40000, 40000)                      # THB a person-year: list, then negotiated twice
town <- c(1400, 1400, 1000); rural <- 2000 - town   # starts a year (5,000 eligible in each)
clinic <- c(0, 0, 32e6)                             # THB: a mobile clinic, bought in year 1
extra <- drug - offset; spend <- 12000 * extra + clinic   # 12,000 person-years over three years
icer <- spend / 4000                                # THB per QALY: 12,000 / 3 = 4,000 QALYs on C
bills <- outer(extra, c(y1 = 2000, y2 = 4000, y3 = 6000)) / 1e6 + cbind(clinic / 1e6, 0, 0)
print(cbind(bills, total = spend / 1e6, pushed_out = spend / k))   # million THB; QALYs
each <- spend / k / 2; net <- cbind(town, rural) * 6 / 3 - each    # 6 person-years per yearly start
z <- (k - icer) / spread; favour <- round(1000 * pnorm(z))         # reruns of 1000 favouring C
print(data.frame(extra, icer, nb = k - icer, z, favour, each, net, total = rowSums(net)))

Output of the run brief_options.log

      y1  y2  y3 total pushed_out
[1,] 100 200 300   600       3750
[2,]  60 120 180   360       2250
[3,]  92 120 180   392       2450
  extra   icer    nb    z favour each town rural total
1 50000 150000 10000 0.25    599 1875  925  -675   250
2 30000  90000 70000 1.75    960 1125 1675    75  1750
3 30000  98000 62000 1.55    939 1225  775   775  1550

The three paying options side by side. It prints the yearly bills, the price per extra QALY, reruns that favour C, net QALYs by district and the QALYs pushed out.

Conditions, triggers and a date to look again

Whatever the board picks, the brief says what would make it look again. These review triggers are agreed before C starts, so nobody has to argue later about whether a change matters. Any paying option can also be conditional: the fund pays while a register of the first patients collects evidence, and the triggers decide when the board looks again. For C the brief lists five:

more than 2,000 starts in a year;
a yearly bill above 200 million THB;
a new drug price;
rural starts below 600 a year;
a complication risk in the province's register far from the trial's 12.12%.

It also names who does what and when. The fund's pharmacy team negotiates the price, and the district clinics report starts every quarter. The board reviews C at the end of year 2, before the year-3 budget is set. Year 3 is where option 1 breaks the budget.

The methods go in a one-page appendix that a reader can check without the model. It lists every input with its value and source: 10,000 eligible, 2,000 starts a year, the drug at 60,000 THB and 10,000 THB saved. It adds 1.00 QALY per patient, the 160,000 THB threshold and a standard deviation of 40,000 THB. It ends with the main assumptions and how each number in the brief was calculated.

The appendix follows the published checklist of what a health economic evaluation should report, CHEERS 2022 (Consolidated Health Economic Evaluation Reporting Standards) [3]. The brief itself cites only the few sources behind its key numbers.

The brief, rewritten

Put together, the analyst's one paragraph becomes two pages in the order the board needs them. Every number from the draft is still there. What changed is the order, the options, and one plain sentence under each number about what it means and what it does not.